The agent-driven video recipes (heygen video create) went out untagged, so
media-use avatar / image-to-video usage landed as generic cli traffic and
wasn't attributable to media-use. Add --headers "X-HeyGen-Client-Source:
media-use" to the generating recipes (persistent flag, allowlisted by the
CLI) so those videos carry client_source in master_video_table meta and show
up in the API dashboards alongside the TTS path (#2365) and CLI (#2368).
Send `X-HeyGen-Client-Source: media-use` on every media-use HeyGen API
request (both auth types, via heygenAuthHeaders + the heygenJSON transport),
so backend billing meta can isolate media-use consumption from other free
TTS and avatar-video usage. Unconditional of auth type — a paying user's
media-use call is still media-use — unlike the OAuth-only cli-source header
that gates the free allowance.
Keep the image-to-video recipe; remove the broader capability table (photo
avatar, digital twin, cinematic, translate, lipsync, ai-clipping, voice). It
drifted toward the API-reference surface OP1 says media-use shouldn't carry, and
--request-schema already self-documents each command. Fix the now-dangling
'(below)' pointer in the image-to-video note.
* docs(skills): add cloud render + variables to CLI skill, media-use generative use cases
The hyperframes-cli skill only documented self-managed AWS Lambda rendering; the
zero-infra HeyGen-hosted `cloud render` path (a real, shipped command with its
own docs page) was absent from every skill, so agents never surfaced it.
- hyperframes-cli: add `cloud` to the frontmatter verb list + entry point; new
Cloud and Variables sections; new references/cloud.md distilled from
docs/deploy/cloud.mdx; routing + workflow rows.
- media-use: add image-to-video recipe (heygen video create type:image) plus a
table of other HeyGen generative use cases (photo avatar, digital twin, video
translation, lipsync, voice design) in references/operations.md; surface them
in the SKILL coverage/provider rows.
- Sync README + docs/guides/skills.mdx catalog entries to mention cloud render.
* docs(media-use): point HeyGen generative use cases at --request-schema
Verified against the installed heygen CLI (v0.3.0): no capability gap that would
need the raw API. Replace hardcoded body-field lists with a pointer to
`heygen video create --request-schema` (self-documenting, can't rot), correct
the image-to-video motion_prompt/expressiveness support, and add the
cinematic_avatar, ai-clipping, and photo-avatar creation paths.
* fix(skills): correct media-use manifest hash (clean-tree regen)
The prior regen was polluted by the gitignored skills/media-use/eval-report.html
(a suppressed mv error left it present), so the committed media-use hash didn't
match a clean checkout. Regenerate with no untracked artifacts present.
* feat(cli): coordinate-frame layout findings in check
Four production compositions shipped with 100-600px layout drift, each a
different coordinate-frame confusion the check graded info or missed
entirely: viewport pixels written as container left/top, gsap x/y
treated as absolute position, a -350px margin fighting flex centering,
and stage-relative path coords drawn into a nested SVG.
Three new layout findings close the class:
- positioned_out_of_parent: an absolute/fixed element rendering mostly
outside its positioning ancestor (warning) — the parent needs no
overflow clipping, which is what let container_overflow miss it.
- box_out_of_canvas: a painted panel breaching the canvas (warning) —
text is canvas_overflow's, media is frame_out_of_frame's, painted
boxes were nobody's.
- connector_detached: a connector path whose endpoints land far from
every anchorable element (warning) — measured coordinates drawn into
an SVG with a different origin.
canvas_overflow additionally promotes from info to warning when held
across samples AND the breach exceeds 5% of the canvas.
All three are persistence-tiered and respect data-layout-allow-overflow.
Verified against the four incident compositions: every one now surfaces
its drift as held warnings (previously: info or silence).
* fix(cli): harden coordinate-frame findings against review false positives
Reworks all three findings after two-lens review (adversarial FP hunt in
real Chrome + maintainer pass):
- escaped_container (was positioned_out_of_parent): uses offsetParent
(transform-aware, skips fixed-as-canvas), exempts fully-detached
callouts within an attachment allowance while still flagging
touching-but-mostly-outside drift.
- panel_out_of_canvas (was box_out_of_canvas): paint alone qualifies
(flat solid panels were a false negative), fully off-canvas rects are
parked entrances and stay silent, pointer-events:none marks decorative
layers, hero-sized breaches warn while small bleeds stay info.
- connector_detached: endpoints via getPointAtLength + getScreenCTM
(viewBox, preserveAspectRatio, group transforms, every command type),
defs/marker/clipPath subtrees skipped, word-boundary connector naming,
containment tier limited to opaque non-ancestor targets (a text-bearing
wrapper contains its own diagram's endpoints).
- canvas_overflow promotion requires partial visibility — a fully
off-canvas rect is a parked entrance, not drift.
Verified: the four incident compositions still surface their drift as
held warnings; the review's false-positive repros (fixed HUD, callout,
parked entrance, corner bleed, marker arrowheads, g-transform and
viewBox-scaled connectors) are clean at warning level. Docs and the CLI
skill reference now describe the coordinate-frame findings.
* fix(cli): panel ownership is geometric — direct-text panels were a silent false negative
A painted panel whose direct text stays in-bounds while its box breaches
the canvas produced neither finding: canvas_overflow measures the text
range and panel_out_of_canvas skipped every own-text element. Skip the
panel finding only when the element's own text ALSO breaches (that
geometry belongs to canvas_overflow); pin the message/fixHint wording of
all three findings with positive assertions; document the SVG-internal
anchor blind spot.
* fix(cli): classify panel decoration by paint kind, not pointer-events
pointer-events:none exempted the framed-painting incident's gold frame
layers — hero content that happens to disable hit-testing. Decoration is
now gradient-only paint (spotlights, textures, vignettes); url() images,
solid fills and borders are content regardless of pointer-events.
* fix(cli): add fixHint to the test-local AuditIssue shape
* fix(cli): gradient stops decide content vs decoration; ownership matches canvas_overflow's tolerance
A gradient with any solid stop (alpha >= 0.6) is content — heroes and
cards painted with linear-gradient were invisible under the blanket
gradient exemption; all-translucent stops (spotlights, vignettes) stay
decoration. The text-ownership check now uses the audit tolerance that
canvas_overflow itself fires at, making the contract strict-mutex: any
text breach past that tolerance cedes the element, so a shallow 20px
text breach no longer double-reports.
The `hyperframes feedback` convention only prompted for a free-text
`--comment` "with the failing composition pattern and what you tried".
Agents dutifully filed vague reports (blank CJK text, mid-run exit, 4K
timeout) with no error string, no failure-mode, and — critically — no
published composition, so none could be reproduced or root-caused.
Two additions to the CLI skill:
- Lead bug reports with `--file-issue` (+ `--dir`), which publishes a
minimal repro of the project to a public URL. A comment alone almost
never lets a maintainer reproduce; the composition is what does.
- Give the `--comment` a concrete bug checklist: exact error string
verbatim + whether output was produced / fell back / hard-exited; the
isolated trigger; exact command + HF_*/PRODUCER_* env; frame/timestamp +
visual defect. Drop the "repeat env" ask (the CLI already attaches it).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(hyperframes-media): surface the real reason a TTS line failed
synthesizeHeygen() swallowed every failure into a bare { ok:false }: a thrown
HTTP error (e.g. 402 plan_upgrade_required from heygenJSON) was caught and
discarded, a missing audio_url / failed audio fetch / failed transcode all
returned nothing. audio.mjs then logged 'TTS failed — omitted' for every line
with zero detail, so the actual cause took a hand-rolled repro to find.
Each failure path now returns an { error } string (the caught message, the HTTP
status, or the specific stage that failed), and audio.mjs appends it to the
anomaly. The subprocess providers (elevenlabs/kokoro) get the same treatment via
a shared synthResult() helper. synthesizeHeygen takes an injectable deps arg so
the failure paths are unit-tested (thrown 402, non-ok fetch, missing audio_url).
* fix(media-use): report wav transcode failures accurately
* chore: regenerate skills manifest
* fix(product-launch): keep media out of frame subcompositions
* fix(product-launch): hoist approved frame videos at assembly
* fix(product-launch): format and refresh media contract
* test(product-launch): harden approved video hoist
* chore: refresh product-launch skill manifest
* fix(product-launch): validate and sanitize hoisted video attrs
* fix(product-launch): allowlist hoisted video attributes
Scaffolded projects' npm run check now invokes the single check command
instead of chaining lint, validate, and inspect (three Chrome boots
become one). The CLI skill, its correctness reference, the entry skill's
capability map, README/docs catalog rows, the Mintlify CLI page (new
check section, deprecation banner on inspect), template CLAUDE/AGENTS
(byte-identical), root CLAUDE/AGENTS, and every creation-workflow skill
that taught the old sequence all point at check. snapshot keeps its
standalone sections; validate/inspect stay documented as deprecated
aliases with their check equivalents.
snapshot --zoom <selector|x,y,w,h> + --zoom-scale (default 3) crops via
Puppeteer clip at raised deviceScaleFactor — density changes, layout
never does. Selector resolves per frame with 24px padding; no match is
a loud error, and a frame whose clamped region is a sliver (element
collapsed or animated off-canvas) is skipped with a stderr note rather
than written as a useless few-pixel image.
check --snapshots additionally writes finding-NN-<code>.png crops for
error findings with bboxes (cap 12, deterministic re-seek in a second
session) and draws labeled annotation boxes on overview frames via a
transient overlay injected only after audits complete. Skill reference
gains the zoom workflow: check reports a finding, zoom into it, fix,
re-check.
- SSOT: the three assemble-index.mjs BGM fallbacks now import
bgmDefaultVolume() from media-use's bgm.mjs instead of duplicating the
0.12/0.9 literals (both reviewers). The cross-skill relative import
matches the existing dependency (each workflow's audio.mjs adapter
already resolves ../../media-use/audio/scripts/audio.mjs).
- STATUS_ROLE_KEY: extended with info|neutral|alert|caution|critical —
same hue-carries-meaning class as the original set (all 3 copies).
- bgm.md: phrase the default as bgmDefaultVolume()/BGM_BED_VOLUME with
"currently 0.12" so the prose survives future tuning.
- fetch-pr.mjs: drop dead mergeCommit field from the gh pr view FIELDS
list (version resolution uses mergedAt only).
- music-to-video assemble-index.mjs: comment documenting why its BGM
stays at 0.8 under VO — music is the content there, not a narration
bed, so the explainer pipelines' 0.12 default deliberately does not
apply.
Not changed: pickAccent's chroma fallback — both call sites pass keyless
capture palettes (tokens.json hex lists), so no status-role keys exist
to filter on; the keyed preset path goes through semanticColors and the
build-frame remix, which this PR already fixed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Behavior fixes surfaced by the prompt-guide validation campaign (Tier 1+2
of the upstream bug list; Tier 3 tracked in #2107). Split out from the
doc-only updates, which follow in a separate PR.
- BGM level: default bed volume under narration was 0.8 linear (~-2 dB,
~16 dB too hot vs voice). Now 0.12 (~-18 dB) via shared
bgmDefaultVolume() in media-use bgm.mjs + assemble-index fallbacks in
faceless-explainer / pr-to-video / product-launch-video. Explicit
volume still wins; silent-film 0.9 and music-to-video unchanged.
Adds bgm.test.mjs (3 cases); bgm.md reference updated to match.
- Caption accent: semanticColors() ranked accents purely by chroma, so a
preserved status red (#dc2626) outranked the brand accent and captions
highlighted in error-red. Status-keyed colors now excluded via shared
STATUS_ROLE_KEY regex consumed by both tokens.mjs and build-frame.mjs
(all three skill copies kept in sync).
- Voice threading: workflow SKILL.md Step 3.1 blocks now instruct
choosing the narration voice from the user's ask and passing
--voice <id>; previously "a male voice" was silently ignored and the
default (Marcia/am_michael) always won.
- fetch-pr shipping version: MERGED PRs get best-effort shipped_version
+ version_source in pr.json (first release published at/after merge,
else default-branch package.json marked unreleased); ingest surfaces
it as a 'Shipped in:' brief line; story-design.md forbids inventing
versions when absent.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two real incidents this week had agents skip /figma entirely and drive
Figma via raw MCP tools (get_metadata/get_screenshot/get_design_context)
when a figma.com URL landed inside a creation-workflow skill. Root cause:
none of the creation workflows mention Figma at all, and the only routing
table that does (/hyperframes) is skipped whenever a workflow is invoked
directly rather than through the entry router — which is the common path.
Going raw loses real infrastructure the CLI/skill guarantees: sanitizeSvg()
before freezing (raw-fetched SVGs are unsanitized), .media/manifest.jsonl
provenance (no cache-hit, no version tracking), and brand-token var()
binding (colors bake as literals, so a later Figma brand change can't
propagate without a full re-import).
Added a "figma source" callout to every creation workflow that could
plausibly receive a figma.com link (product-launch-video, website-to-video,
general-video, motion-graphics, slideshow), plus a defense-in-depth line
in /hyperframes's own routing checklist. The fix lives in the workflows
themselves so it doesn't depend on the entry router being consulted.
Co-Authored-By: Claude Opus <noreply@anthropic.com>
* feat(media-use): usage visibility — shared telemetry identity, miss log, resolve --stats
- U6: join the CLI/studio telemetry identity — read the shared install id from
~/.hyperframes/config.json (seed if absent) instead of a media-use-only
~/.media/anon-id, and $identify to the HeyGen account (email/username) once
per run on sign-in. One PostHog person across surfaces; pseudonymous before
sign-in, account-linked after. Event properties stay coarse (no intent/paths).
- U1: one-time first-run disclosure to stderr + Privacy section in SKILL.md;
honors DO_NOT_TRACK / HYPERFRAMES_NO_TELEMETRY.
- U2: persist resolve misses to ~/.media/misses.jsonl (local → intent kept;
the media_use_resolve_miss telemetry event stays intent-free).
- U3: `resolve --stats` (+ --days) — local usage report over .media/ + ~/.media
(volume by type, source/provider/via split, hit-rate, top missed intents,
global-cache size/reuse); human + --json.
- U4: reproducible PostHog dashboard definition (references/telemetry-dashboard.md).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(media-use): address #2113 review — shared notice state, legacy id migration, stats robustness
- Notice-shown state now lives in the shared ~/.hyperframes/config.json
(config.telemetryNoticeShown, the CLI's own field) instead of a media-use-only
~/.media marker — so shared-identity users see the first-run notice once per
person, not once per tool.
- Migrate a pre-existing ~/.media/anon-id into the shared config on upgrade, so
media-use-only users keep their PostHog persona instead of resetting.
- buildStats: --days only windows on a positive finite value (negative/NaN → all
time, not an empty report); dropped the top-level catch that masked a real
error as an all-zero "no usage" report (sub-reads are individually guarded).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rames's inline findings on #2112:
- forbiddenError now RETURNS in every branch (BAD_TOKEN no longer throws
inside) so the caller's single throw covers all cases — no mixed
throw/return contract for a future wrapping caller.
- retryAfterMs capped at 60s: a spec-legal Retry-After: 3600 no longer
silently blocks the CLI for an hour before RATE_LIMITED.
- asset ref gathering extracted to gatherAssetRefs() and made URL-safe:
bare fileKey:nodeId tokens comma-split, but a figma URL with commas in
its query (multi-select node-id=1:2,3:4) is kept whole.
- Documented in SKILL that 429 retry lives in the shared request path, so
EVERY read endpoint retries (not just asset) — blast-radius note the
reviewer asked for. variables intentionally still retries: its fallback
is REQUIRES_ENTERPRISE-only, and a 429 there is transient, not a gate.
Tests: retry-cap (3600→60000), non-styles endpoint retry, gatherAssetRefs
URL-vs-bare split. client 24, cli asset 11.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(media-use): fast heygen CLI onboarding — actionable diagnostics, --doctor, free-usage framing
media-use resolves bgm/sfx/image/icon (catalog), voice (TTS), and avatar video
through the heygen CLI — the free-usage path. Agents hit a dead end when it's
missing/unauthed. This guides them to install it fast, at the point of need.
- Centralized actionable diagnostics (lib/heygen-cli.mjs): every heygen-backed
resolve, on failure, prints the exact fix on stderr — not-installed (curl
install one-liner), not-authenticated (heygen auth login), outdated (heygen
update). Routed through heygen-search + voice-provider. stdout stays clean JSON.
- resolve --doctor preflight (human + --json): checks heygen present/version/
auth, ffmpeg, ffprobe, node, a fix per gap. Exit 0 unless ffmpeg missing.
- SKILL reframe: install-first callout; heygen as the free-usage gateway for
bgm/image/voice/avatar-video; removed the false "degrades gracefully" claim.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(media-use): address #2065 review — classifier blocker, doctor contract, telemetry
- Blocker: classifyHeygenError no longer treats a bare "not found" as CLI-missing
(a stale voiceId → "voice not found" was sending users to reinstall a working
CLI); keep only ENOENT + "command not found". Regression test added.
- 401 now matches \b401\b, not any "401" substring (request IDs no longer misread).
- --doctor: top-level ok requires ffmpeg AND ffprobe (matches SKILL.md); emits
media_use_doctor_run telemetry; auth status queried with --json + JSON-only
parse; auth timeout softened (network issue, not a false "unauthenticated");
node version gated on >= 18; version-without-semver labeled, not silently green.
- Nits: install cmd uses && ; dropped the runResolveStatus alias.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(media-use): require OAuth-capable heygen CLI (v0.3.0), fix auth-status probe
E2E against the live free-usage backend surfaced three issues:
- HEYGEN_MIN_VERSION was 0.1.6, but that CLI can't use OAuth ("heygen-cli can't
use OAuth yet") — free usage needs >= v0.3.0. Bumped the floor; --doctor now
also nudges `heygen update` when a newer stable exists (always-latest).
- Onboarding pointed at `heygen auth login --key` (API credits / billing); the
free path is `--oauth` (subscription/free credits). Fixed install + auth
guidance and SKILL.md accordingly.
- `heygen auth status --json` is an unknown flag on v0.3.0 (JSON is the default
output) — the added --json broke auth detection. Dropped it; verified
--doctor reports authenticated on a real free (OAuth) account.
Tests assert against the exported message constants instead of brittle literals.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(media-use): address #2065 review nits — one root cause on old CLI, floor policy
- --doctor skips the auth check when the version check fails (below v0.3.0): an
old CLI's auth probe fails for the same root cause, so users no longer see two
errors ("outdated" + "not authenticated") — one root cause, one fix.
- Comment links the auth-status probe's JSON-default assumption to
HEYGEN_MIN_VERSION >= 0.3.0 so the floor isn't silently lowered later.
- SKILL.md states the uniform v0.3.0 requirement (nudged even for API-key use).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(media-use): doctor prints one heygen row per fact
The 'heygen on PATH' and 'heygen version' checks both rendered their
detail as `heygen v0.3.0`, so --doctor printed two byte-identical green
lines. Make the PATH row report presence ("heygen found on PATH") and let
the version row own the version string — one row per fact, no duplicate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(media-use): use CLI free HeyGen usage
* fix(media-use): address #2027 R1 nits — gate cli-source header to OAuth, export origin constant
- X-HeyGen-Source is now sent only on OAuth (Bearer) requests, not API-key ones —
the backend ignores it for API-key traffic (normal billing), so it was dead
metadata there. buildAuthHeaders + heygenAuthHeaders + tests updated.
- Export HEYGEN_CLI_ORIGIN_HEADER ("X-HeyGen-Client-Origin") for future cli:<origin>
consumers.
- Document the deliberate paid/X4 confirm-before-call decision on heygen.tts.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* refactor(cli): drop unused origin-header export, dedup auth-client tests
Fallow flagged 5 findings on this PR:
- major: HEYGEN_CLI_ORIGIN_HEADER was exported but never emitted or
imported — speculative dead code ("future consumers"). Remove it; a
real consumer can add the constant when one exists.
- 4x minor duplication in client.test.ts: fold the repeated
`.rejects.toSatisfy(auth-code)` assertion into expectAuthCode(), and the
repeated try/catch scrubbed-message assertion into expectRejectionMessage().
No behavior change; auth/client tests still 17/17.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extends the scope+retry work from the figma bug-bash (valid report:
9-bugs-with-repros; the skill-not-used report was discarded).
- 403-body parse (bug 4): figma returns 403 {"err":"Invalid token"} for bad
PATs (NOT 401), and 403 {"err":"Invalid scope(s)… requires X"} for missing
scopes. get() now reads the body: "Invalid token" reclassifies to BAD_TOKEN
with re-mint advice; a scope body surfaces figma's own diagnosis verbatim;
else falls back to the endpoint's scope hint. Reads both err and message
(variables endpoint uses message). One fix, honest messages for bugs 1/4/9.
- Batch asset fetch (requested): figma asset accepts multiple refs
(space-separated or comma-joined) of one file and renders them in a SINGLE
/v1/images call via new client.renderNodes — figma's documented per-minute
rate-limit workaround. runAssetImport delegates to runAssetImportMany;
cache-checks per node, batches only the misses, one index.md regen.
- NO_TOKEN box (bug 8): errorBox indented only the first hint line, mangling
the numbered setup list. Indent every line; single-line hints unchanged.
Verified live: 3 refs -> 3 imports -> 1 request; bad token -> BAD_TOKEN not
scope advice. Client suite 22, cli figma 33.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
## What
Fixes five reported false-positive/false-negative patterns in the WCAG contrast audit (`hyperframes validate --contrast`):
1. **SVG fill vs. text color** — foreground read from CSS `color` instead of SVG `fill`.
2. **Cross-component color bleed** — background estimate bleeds into a neighboring panel/layer.
3. **Backdrop-filter glass text** — background estimate misses the blur/tint and reads the raw backdrop.
4. **Partially-overlapping translucent decoration** — a decorative shape inside or partly touching the text's bbox goes undetected.
5. **Solid-fill pill/button** — investigated, did **not** reproduce; already handled correctly by the existing own-background ancestor walk. Not touched.
## Why
The audit estimated an element's background two ways:
- foreground: always `getComputedStyle(el).color` — wrong for SVG `<text>`/`<tspan>`, which is painted via `fill`, an independent CSS property.
- background: a 4px pixel ring sampled just **outside** the text's bounding box, with a fallback to an ancestor's opaque `background-color` for solid pills/buttons.
The ring is a proximity heuristic. It's wrong whenever what's immediately outside the text differs from what's actually behind it:
- text near the edge of its own panel, with a differently-colored sibling panel/layer just past the bbox — the ring samples the neighbor.
- a `backdrop-filter: blur()` glass panel sized only a couple pixels larger than the text — the ring exits the panel into the raw, unblurred, untinted backdrop.
- a translucent decoration that only partially overlaps the ring, or sits entirely **inside** the bbox — invisible to the ring regardless of size.
## How
**SVG fill (#1):** elements inside an `<svg>` (`el.ownerSVGElement`) now prefer the computed `fill` when it resolves to a solid `rgb()`/`rgba()` color, falling back to `color` for paint values that aren't a plain color (`none`, `context-fill`, gradient/pattern refs).
**Cross-comp bleed / glass blur / partial decoration (#2–#4):** replaced the ring-sampling + own-background-ancestor-walk heuristic with a two-phase capture:
1. `__contrastAuditPrepare()` walks the DOM, computes each candidate's foreground (unchanged logic from #1), and **hides that element's own text paint** (`color`/`fill` → `transparent`, layout-neutral — no reflow).
2. The caller takes **one** screenshot with the glyphs invisible (same number of screenshots as before — just moved after the hide instead of before it).
3. `__contrastAuditFinish(imgBase64, time, candidates)` restores the original paint immediately, then samples the **real composited pixels directly inside each element's own bbox** — no proximity heuristic needed, since these are the exact pixels that were behind the glyphs.
This is a real architectural change to `contrast-audit.browser.js`'s calling contract (single `__contrastAudit` → `__contrastAuditPrepare`/`__contrastAuditFinish`), with `validate.ts`'s `runContrastAudit` updated to match, including a try/finally restore-safety-net so a mid-loop screenshot/decode failure can't leave a later sample auditing a page with stale hidden text.
Mirrored the identical change in `skills/hyperframes-creative/scripts/contrast-report.mjs`, which duplicates the same DOM-walk/sampling logic (not just the WCAG math). There, the **visible** frame for the human-facing overlay image still comes from the producer's normal `captureFrameToBuffer` path (unchanged); only the **background-sampling** capture is a plain `session.page.screenshot()` taken after hiding text — deliberately bypassing `captureFrameToBuffer`, whose static-frame dedup cache knows nothing about the DOM mutation and would hand back a stale pre-mutation buffer.
**Solid-fill pill (#5):** reproduced a rounded pill/button with a busy page background outside it. The existing own-background ancestor walk already resolves the pill's declared `background-color` correctly regardless of the rounded corners — confirmed via repro, both before and after this change report the identical (correct) result. No fix needed; left untouched, and this case is covered by the new architecture too (would give the same right answer even without the ancestor-walk fallback).
Added `packages/cli/src/commands/contrast-sample.ts` (mirroring the existing `contrast-bg.ts`/`contrast-fg.ts` pattern) hosting the pure sample-rect/grid-point computation, unit tested — the browser-injected scripts can't import it directly, so it's kept in sync by hand, same convention as the rest of this file.
## Test plan
- [x] Unit tests: `contrast-fg.test.ts` (SVG fill resolution), `contrast-sample.test.ts` (sample-rect clamping/degenerate cases), plus the full `packages/cli` suite (1424 tests) passes, including an updated `layout-audit.browser.test.ts` case that called the old single-function `__contrastAudit` API directly.
- [x] Manual verification — standalone `puppeteer-core` harness against real `chrome-headless-shell`, one minimal HTML fixture per pattern, comparing the audit's reported ratio/verdict against a hand-constructed ground truth:
- **SVG fill**: `fill:white` / no `color` on black bg → before: `fg=rgb(0,0,0)` ratio `1:1` (false FAIL); after: `fg=rgb(255,255,255)` ratio `21:1` (correct PASS).
- **Cross-comp bleed**: text on a black sibling highlight box 2px larger than the text, white page bg outside it → before: `bg=rgb(255,255,255)` ratio `1.23:1` (false FAIL); after: `bg=rgb(0,0,0)` ratio `17.14:1` (correct PASS).
- **Glass blur**: black text on an 18%-white-tinted `backdrop-filter: blur(14px)` panel over a yellow/blue gradient, panel only ~2px larger than the text → before: `bg=rgb(0,64,255)` (raw gradient color, blur/tint completely missed) ratio `3.18:1` (false FAIL); after: `bg=rgb(159,160,165)` (correct blurred/tinted blend) ratio `8.05:1` (correct PASS).
- **Partial decoration**: text 92%-covered by a translucent white badge on a dark bg → before: `bg=rgb(16,16,16)` (ring never touches the badge, which sits entirely inside the bbox) ratio `17.45:1` (false PASS); after: `bg=rgb(171,171,171)` (correctly detects the badge) ratio `2.11:1` (correct FAIL).
- **Solid pill sanity**: unaffected — `bg=rgb(10,10,10)` ratio `19.8:1` before and after.
- [x] End-to-end: ran the actual `hyperframes validate --contrast` CLI command (via `tsx src/cli.ts`) against a real scaffolded project containing all 4 patterns simultaneously — only the genuinely-failing case (the 92%-covered decoration) is reported (`1.09:1`, need `3:1`); the cross-comp-bleed, glass-blur, and solid-pill cases are correctly silent. A second vanilla scaffold with plain white-on-dark text produces zero false positives.
- [x] `oxlint`, `oxfmt --check`, and `tsc --noEmit` all pass on the changed files.
Two bugs from live figma-integration use:
1. `tokens` styles fallback 403s on non-Enterprise. /v1/files/:key/styles
needs library_content:read — a scope the setup docs and the generic
FORBIDDEN message both omitted, so the user saw "missing a read scope"
with no way to know which. Each endpoint now carries a scope hint; the
403 names the exact scope (styles → library_content:read). Setup text and
skill scope list updated to include Library content: Read-only.
2. `asset` (and every per-node component render) had no 429 handling — the
message said "back off and retry" but the client didn't. Two imports in
a row tripped the per-minute limit and hard-failed. get() now retries 429
with exponential backoff, honoring Retry-After when present, before
surfacing RATE_LIMITED after maxRetries (default 3). sleep is injectable
so tests don't wait.
Batch multi-node asset syntax (the documented /v1/images comma-ids rate
workaround) is a separate enhancement — retry makes the reported failure
self-heal, including the many-node component path.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A figma text node whose box is shorter than its line-height carries
vertically-trimmed (cap-to-baseline) bounds. The mapper positioned the box
at those bounds but let the browser lay glyphs with half-leading, pushing
them ~6px low on a 70px font (glyph-centroid measurement against figma's
own render: +9.1px vs figma's +3.4px inside the same pill). Emitting
text-box-trim: trim-both / text-box-edge: cap alphabetic reproduces the
trim in the render engine; post-fix centroid agrees within 0.4px and the
motion verifier's min window score improved 20.3 -> 25.3dB. Trim applies
only to single-line trimmed text; boxes matching their line-height are
untouched.
Skill: component imports now include a static fidelity self-check step
against figma's PNG export of the same node.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- motionContextToDocs: escape regex metacharacters in arrayAfterKey /
scalarAfterKey key interpolation (safe today for \\w+ keys; now safe for
any future caller), and document balancedBlock's no-strings invariant.
- verify-motion.mjs: execSync shell string -> spawnSync with array args
(JSON.stringify is not shell escaping); verifier re-calibrated unchanged
(faithful render still PASS at min 20.30dB).
- command-failure-tracking: rebase folded the group-delegation skip into
upstream's recursive wrapCommand (HF#2033) — leaf commands now assert
their own flag tables, so `figma component --namee` is rejected at the
leaf while `--name` passes the group; heuristic invariant documented.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two guarantees so figma-motion imports can't drift from the design again:
- motionContextToDocs(): raw get_motion_context response -> MotionDoc[],
in code. Parses the motion.dev snippets (the reliable encoding; the CSS
snippets stretch durations and can disagree), strips loop-wrap tail
keyframes (sub-ms segments at the window end are the loop reset, not
authored motion), preserves bezier eases verbatim. Fixture test uses the
verbatim response from a real Motion timeline whose translation was
frame-validated against Figma's own export_video render.
- skills/figma/scripts/verify-motion.mjs: mandatory post-render gate.
Compares motion-energy deltas between the render and the export_video
ground truth so static import fidelity cancels out and the score
isolates choreography. Calibrated on a faithful translation (min 20.3dB)
vs a diverging one (min 5.0dB); threshold 15dB.
The skill's Motion step now routes through both: no hand transcription,
no unverified completion.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Field lesson from translating a real Motion timeline: the two returned
encodings window durations differently, and keyframes at times ~0.9999
are loop-wrap resets, not authored motion. Hand-normalizing across
encodings and inventing visible returns produced a render that diverged
from Figma. The skill now mandates verbatim single-encoding translation,
wrap-via-repeat, and a frame-grid comparison against export_video ground
truth before completion.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* feat(media-use): color grading — grade/lut resolve, smart-grade, grade-compare CLI
Add color grading to media-use as first-class resolve types plus a faithful
comparison command. All local, offline, deterministic — no model, no GPU.
- resolve -t grade / -t lut: produce a data-color-grading block (or a frozen
.cube). Look cascade: core preset (no file) -> bundled .cube library ->
parametric buildCube. Emitted .cube is Rec.709 and validated against core's
colorLuts constraints (LUT_3D_SIZE <= 64) before it is frozen.
- smart grade (grade --for <media>): ffmpeg signalstats -> adjust suggestion
(exposure / contrast / white balance), surfaced with the measured evidence on
stderr as a starting point; never auto-applied.
- hyperframes grade-compare: renders N candidate grades onto a reference frame
through the real runtime shader into one labeled comparison PNG, so an agent
picks a look without opening Studio. Prepends an "original" baseline cell by
default (--no-baseline to omit). Shares the headless-capture pipeline with
snapshot via capture/captureCompositionFrame.
- media-use SKILL: proactive "media opportunity pass" guidance (grounded
signal -> offer, ask once, surface don't mutate).
Verified: media-use 116/116, grade-compare 7/7, snapshot 9/9, lint + format
clean, full build green, comparison renders end to end.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* test(cli): narrow grade-compare baseline assertion off unknown-typed grading
Assert the whole cell via toEqual instead of reaching into .grading.preset /
.grading.lut on the unknown-typed field, keeping the test typecheck-clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* feat(media-use): agent-authored LUTs via --params + validate --from cube; never-read-.cube guardrail
- resolve -t lut / -t grade --params '<json>': build a parametric .cube from
explicit params (bypassing the intent cascade), validate, and freeze in one
step. --intent becomes the optional description. Lets an agent commit a look
it computed itself.
- --from <file.cube> now validates the ingested LUT for lut/grade types and
rejects an invalid/oversized cube (no partial write) — the escape hatch for a
LUT the agent generated with its own code.
- SKILL.md: hard rule to never read a .cube body into context (~size^3 lines,
zero legible signal) — inspect via grade-compare (see it) or cube-validate
(ok/size), read the manifest description for meaning; plus both authoring
paths and the parametric-vs-film-stock ceiling note.
Verified: media-use 116/116, lint + format clean; smokes — --params builds a
valid frozen cube, grade --params returns a lut block, bad JSON and an oversized
--from cube are both rejected with no stray file.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(cli): grade-compare validates referenced LUTs, warns on no-op cells, caps candidates
Bug-bash follow-ups — grade-compare silently accepted bad input:
- Validate LUT *content*, not just existence: each referenced .cube is parsed
with core's parseCubeLut (now exported from @hyperframes/core) and rejected
with a per-cell error ("LUT for \"<label>\" is not a valid .cube: ..."). A
file that exists but isn't a valid cube no longer renders a silent no-op cell.
- Warn on inactive cells: a grading that normalizes to inactive (e.g. a
malformed {lut:12345}) emits a stderr warning naming the cell; the
auto-prepended "original" baseline is intentionally inactive and stays silent.
stdout remains valid JSON.
- Cap candidates at 16 (excluding baseline): over-cap input renders the first N
and reports {truncated:true, total:M} on stdout + a stderr note — no silent
drop, no unbounded giant sheet.
Verified: grade-compare 10/10; non-cube LUT → clear error; {lut:12345} → warning
+ ok; 20 cells → cells=17 truncated total=20; valid runs unchanged. Lint/format
clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* feat(cli): general `hyperframes compare` visual-variant primitive
Generalize grade-compare's "render N variants → one labeled sheet → the agent
looks and picks" loop into a standalone command that works on ANY variation
(font, layout, motion, grade, whole compositions) — the tool never needs to
know what differs.
- `hyperframes compare <path...> [--at <sec>] [--labels a,b,c] [--out] [--cols]
[--json]`: renders each agent-authored composition variant through the real
runtime (captureCompositionFrame) and stitches one labeled comparison sheet +
JSON ({ok, sheet, rendered, variants, truncated?/total?}). 2+ paths required;
caps at 16 with loud truncation. It presents, it does not judge — choosing is
the caller's job.
- Factored the shared "render a labeled set → contact sheet" path so compare,
grade-compare, and snapshot all sit on it (no duplication). grade-compare is
now the first color-specific specialization of this primitive.
- New pathArgs util + contactSheet test; hyperframes-cli SKILL documents compare
as the agent's "see your own renders and choose" primitive.
Verified: 26/26 across compare + grade-compare + snapshot + contactSheet (no
regressions); compare renders 3 variants into one visibly-distinct labeled
sheet; 2+-path error path clean; lint/format clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(ci): green the skills CI — skip ffmpeg tests when absent, oxfmt markdown
The "Test: skills" CI job runs bare `node --test` with no ffmpeg on PATH (by
design — skills tests are meant to be node-builtin-only). The grade-analyzer +
smart-grade tests shell to ffmpeg and were failing there with ENOENT. Guard
them to skip when ffmpeg isn't on PATH; they still run locally / where it is.
Also oxfmt README.md + hyperframes/media-use SKILL.md (the whole-repo
`oxfmt --check .` Format job caught markdown left unformatted by the rebase
conflict resolution).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(ci): skip core-conformance test when tsx is unavailable
The "Test: skills" CI job installs no deps, so the normalizeHfColorGrading
conformance test (which imports core's TS via `node --import tsx`) failed there.
Guard it to skip when tsx can't resolve; runs locally / in the deps-installed
Test job. Completes the skills-CI greening (the ffmpeg guards handled the rest).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(cli): escape grade-compare src double-quotes (CodeQL XSS) + Windows-safe compare test
- grade-compare built `<img src="...">` (double-quoted) with the single-quote
escaper, leaving `"` unescaped — a `"` in the frame path could break out
(CodeQL: incomplete HTML attribute sanitization). Use escapeXml for src.
- compare label test hard-coded POSIX paths that can't match on Windows; assert
the derived labels (the subject); path resolution is covered elsewhere.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* refactor(media-use): generate LUT library from params (drop committed .cube files)
The 3 bundled .cube files were 733 lines each (2,199 total) and were themselves
buildCube output — pure repo bloat. Replace with compact per-look params in
luts/index.json, generated on resolve; add an optional `url` for future scanned
LUTs to be CDN-hosted + downloaded on demand (freezeUrl) instead of committed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* feat(media-use): serve library LUTs from CDN on-demand (static.heygen.ai/luts), params fallback
Looks now carry a CDN `url` (hosted at s3://heygen-public/luts → static.heygen.ai/luts/<id>.cube);
resolve downloads + validates + freezes on demand, like bgm/image. `params` stays
as the deterministic offline fallback (--local-only, or if the download fails), so
resolution is never blocked on the network. Provider prefers url, falls back to params.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
* fix(media-use): address #2041 review — atomic LUT writes, compare telemetry, follow-ups
- Atomic .cube writes: library provider (url + params) and the parametric
generator now write to a .tmp path, validate, then rename, so a crash can
never orphan an invalid .cube at the final path (was validate-after-write).
- track("media_use_resolve") now emits provenance.via (url/params-fallback/params).
- grade-compare + compare: --timeout flag (was hardcoded 5000) and a
media_use_compare event (cells, truncated, total, render_ready_timed_out);
openSettledCompositionPage now surfaces the render-ready timeout.
- compare staging skips node_modules/.git; --for gets an upfront existence check.
- Rec.709 luma comment; HYPERFRAMES_ANALYZE_TIMEOUT_MS override; measured note
uses basename; LUT s3 hosting moved from index.json into luts/README.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H5k87mPZ4d6yiFwcWSb8Vv
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(skills): add brief contract — interaction modes + shared intake fields across workflows
New hyperframes-core/references/brief-contract.md, the shared intake
contract every creation workflow now runs its brief against:
- §1 interaction mode: collaborative (default) vs autonomous, ongoing
vs one-time signals, mode set once and carried forward, and a gate
taxonomy (preference / checkpoint / quality / routing) — autonomous
skips waiting, never verification
- §2 field registry: destination→aspect derivation (feed → 1:1,
Shorts/TikTok → 9:16, else 16:9), message, angle, length, audience,
language, narration — each workflow binds fields as ask or state
- §3 question rules: one round with one question per asked field
(native question UI mandatory when available, recommended option
first with a receipt), never drop a question as inferable, and a
mode legend advertised in the intro text instead of asked
Wired into the surfaces:
- hyperframes router: detect mode at entry, derive aspect from
destination instead of stating 16:9
- product-launch-video / pr-to-video / faceless-explainer: ask/state
binding tables at Step 0; Step 3/6 checkpoint-gate branches
(autonomous posts a heads-up with a preview hint before render)
- website-to-video: local mode definition now defers to the contract
- music-to-video, general-video, embedded-captions, talking-head-recut,
slideshow, motion-graphics: mode semantics wired per gate type
- storyboard-format: new optional 'mode' frontmatter key
- pr-to-video: length tier is a ceiling, not a floor — a one-headline
PR recommends inside the 30–90s sweet spot regardless of diff size
* feat(skills): story spine + mode-first brief across creation workflows
Story — the reverse-iceberg feedback:
- New hyperframes-creative/references/story-spine.md, three rules for
the narrated workflows: the hook speaks the viewer's outcome
language, the value claim lands by beat 2 (implementation is the
footnote of the story, not the spine), and the storyboard is
presented as a proposal — 'This video tells [audience] that
[message]' plus a per-frame why: drawn from narrativeRole
- pr-to-video: feature-reveal reordered promise-first (impact leads,
diff/mechanism follow as evidence); hooks ban file/function names;
fix-explainer, refactor-walkthrough, changelog unchanged
- product-launch-video / faceless-explainer hook rules aligned to the
spine; website-to-video's beat summary gains the echo line + why:;
general-video points at the spine from its plan step
Brief — hardened after live-test drift:
- Mode is now the first question (Collaborative recommended vs
Autonomous), its own round, skipped when the request carries a
signal; autonomous asks nothing further until one final
preview-or-render question before render
- Step 0 rewritten as a literal two-round question script in each
shot-sequence workflow (website-to-video's editorial register,
channel-agnostic); brief-contract.md §3 reduced to invariants so the
procedure lives in exactly one place
* feat(skills): split type minimums by viewing context
typography.md: full-screen viewing keeps body 20px / headline 60px;
in-feed destinations (X / LinkedIn / Instagram — brief-contract's
destination field) scale to body >=32px, headline >=90px, data labels
>=24px. First-pass values, to be calibrated against real renders.
* feat(skills): storyboard proposal as a table + credits close by default
- story-spine § 3: the proposal presents frames as a markdown table
(frame · beat · on screen · why) instead of dense paragraphs; the
three shot-sequence workflows and website-to-video's beat summary
reference the same shape
- pr-to-video: the credits close is now the default ending — every PR
video ends on a contributors frame (committers by commit count, 1-6
avatars), with no taste judgment; the only skip is when no avatar
was fetched, and the user can cut the frame in the proposal
* fix(skills): address review nits on the brief/story contracts
- embedded-captions: the identity procedure now states both sides of the
preference gate inline (user picks; autonomous picks with a stated why)
- website-to-video step-2-brief: note that its mode section is the
workflow's application of brief-contract.md, not a second definition
- brief-contract: resuming a project reads mode from STORYBOARD.md
frontmatter — a recorded mode counts as set, closing the write-only gap
* docs(skills): add a non-code receipts example to the brief contract
Review nit (jrusso1020, #2058): the receipts example in § 3 was
PR-video-shaped only. A destination-shaped example joins it so the rule
reads as workflow-neutral.
* fix(cli,skills): install workflow skills on demand instead of re-pulling the full set
Users report every init re-pulls all 21 skills into ~/.agents/skills
whenever anything is stale or missing - heavy, noisy, and it re-expands
deliberate partial installs.
Split the set into two tiers:
- core: the /hyperframes router + hyperframes-* domain skills +
media-use, which every workflow references structurally. init and
bare 'skills update' keep these (plus anything already installed)
fresh, and never expand the install.
- on demand: the end-user workflow skills (and figma). They install at
trigger time via 'skills update <name...>' - positional names are
the only way update expands an install: one targeted
'skills add --skill <name>' covering only stale/missing targets, a
fast no-op when current, presence-verified after install, exit 1 on
unknown names, and a presence-only degrade when GitHub is
unreachable.
The /hyperframes router now runs 'skills update <workflow>' after
routing and before reading the workflow skill, so a routed workflow is
guaranteed present even on a machine that only has the core set. Each
on-demand skill also opens with the same self-maintenance step (run
'npx hyperframes skills update <name>' silently), so a workflow
triggered directly - without the router - still refreshes itself and
restores any missing core skill before relying on it.
When the manifest is unreachable (offline / rate-limited) the engine
degrades honestly instead of claiming success: named runs presence-check
the request plus a pinned fallback core list (unit-pinned to skills/)
and blind-install whatever is absent; a bare strict update fails loudly
so the 'check || update' chain can't pass while everything stays stale;
init reports the skipped freshness check. --json emits structured
errors on failure paths.
skills check still lists every skill, but exits non-zero only for
stale installed skills, an incomplete core set, or removed leftovers -
workflow skills not yet installed are reported as available on demand.
Bare 'hyperframes skills' (and 'skills add --all') remain the explicit
full-set installs.
Verified end-to-end with a sandboxed $HOME: fresh init installs the 9
core skills only; 'skills update slideshow' adds exactly that skill
(no-op on re-run, exit 1 on unknown names); bare update refreshes
without expanding; a live Claude Code run routed PR-to-video, executed
the router's update step, and the workflow skill appeared before use;
and a second live run triggered an installed workflow directly, whose
opening maintenance step restored a deliberately removed core skill.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs(skills): clarify update-engine contracts + document lazy-install model
- skills.ts: note the UpdateSkillsResult.unknown strict-mode contract,
verifyInstalled's non-strict (warn-not-throw) intent, and that a
partial install stays "refreshed but never expanded" (review nits).
- docs/guides/skills.mdx: add a "Keeping skills current" section covering
the core-eager / workflow-on-demand model and the skills check|update
commands, per the repo's catalog-maintenance rule.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: kiritowoo <295860553+kiritowoo@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Miao Yang <miao.yang@heygen.com>
* feat(media-use): resolve official brand logos via a four-tier cascade
Third-party brand logos (the meeting's 'credibility signals lost' gap)
had no acquisition path: capture only grabs the product's own site
assets, and HeyGen asset search returns generic look-alike icons for
brand queries (0/3 in testing — an X-in-a-circle for LinkedIn). Workers
could only fake a mark or drop it.
New resolve type 'logo', four tiers verified by a 54-brand stress test
(100% cascade hit across dev tools / big tech / non-tech / CN brands):
- svgl — official full-color vector SVGs + wordmark variants (40/54
first-hits); search is substring-based, so entities pass through
alias normalization (nextjs → 'next.js', aws → 'amazon web services')
- simple-icons (pinned CDN build) — official monochrome glyphs; catches
the long tail (nike, visa, toyota, wechat, bytedance)
- github org avatar — known-org map only; a brand name is not a GitHub
login, guessing risks same-named personal accounts
- domain favicon (DuckDuckGo ip3) — small-raster last resort; sub-500B
responses are DDG's placeholder and rejected; frozen with a low_res
provenance flag (chip-size use only)
logo joins the icon/image equivalence group (typesMatch) and the
images/ subdir, so entity cache hits interop with figma-imported marks.
A total miss falls through resolve's normal failure path — no special
casing. HeyGen search stays the icon provider; it is deliberately
absent from the logo cascade.
Docs: media-use gap/types/providers tables + example; the five
workflow banners now cover logos (catalog claim kept for media, 'from
their official sources' added for logos); product-launch story-design
and motion-graphics logo-reveal point at the new type; catalog
surfaces (CLAUDE.md / README / docs) updated in lockstep.
Verified: 19 unit tests + coverage row green; live smoke across all
four tiers (linkedin→svgl, nike→simple-icons, heygen→github.avatar,
amazon→favicon) plus a fabricated brand exiting 1 on the default miss
path. oxlint + oxfmt clean.
* test(media-use): sanction the four logo providers in the registry allowlist
svgl / simple-icons / github.avatar / favicon.ddg join the sanctioned
list — the logo cascade added in the previous commit. Full lib suite
95/95 green.
* test(media-use): gate the logo cascade behavior in CI + single-fetch favicon tier
Review follow-ups (miga-heygen, jrusso1020 on #2061):
- Eight mocked-network tests pin what the manual 54-brand stress test
only asserted: descriptor shape, alias retry (svgl non-array payload
→ next query, simple-icons 404 → next slug), network-error → null
fallthrough, the sub-500B placeholder rejection, github's
no-guessing (zero fetches for unmapped entities), and the real
cascade order landing tier by tier under a mocked network.
- faviconSearch now hands its verified bytes over as a local file, so
the freeze step copies instead of re-downloading — one round-trip,
and the size check is authoritative over what gets frozen.
- The header's hit counts are labeled as a stress-test snapshot, not a
live invariant.
Full lib suite 103/103; live smoke re-verified (amazon → favicon.ddg,
frozen .ico).
Brand-loop live test (SDS duplicate, plans/figma/brand-loop-test-plan.md)
proved the recolor chain end-to-end and surfaced three gaps:
- runtime now defines every declared composition variable as a CSS
custom property (document root at init + scoped sub-comp hosts in the
loader), so imported var(--slug, literal) fills resolve live — without
this the frozen literal always won and variable-driven rebranding
could not propagate. Slug kept byte-compatible with the figma
importer (parity test). render --variables overrides win.
- figma component --name: variant frames are often all named
'Platform=Desktop' and slug-collided across imports.
- imported fragments carry data-hf-snippet and the project linter skips
composition-root rules for them.
- /figma skill documents the field-tested non-Enterprise tokens path
(MCP get_variable_defs joined with REST boundVariables ids).
Shared-helper extractions (injectScopedStyles, flattenedRoot module,
parseHostVariableValues, rasterizeFallback, shapeCss) satisfy the
dedup/complexity audit the runtime changes tripped.
Validated live: brand-loop renders purple from the attribute alone (no
manual :root); 118 figma + 662 runtime/compiler + 331 lint tests green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
When codex is a shell alias (e.g. codex → /Applications/Codex.app/...), a
spawned subprocess's PATH lookup can't see it, so media-use reported the bare
"codex CLI not on PATH" — which reads as "binary missing" and sent people
hunting for a bad install. Spell out the actual cause (aliases aren't visible
to child processes) and the one-line fix (symlink the real binary onto PATH).
No behavior change; clearer agent/user guidance only.
* fix(media-use): codex gate misfires as 'not logged in' when piped
codexUnavailableReason() gated generation on parsing `codex login status`
stdout, but that command prints 'Logged in using ChatGPT' to stderr and
exits 0 — so the piped stdout media-use captures (execFileSync returns
stdout only on success) was empty, and the gate falsely reported 'not
logged in'. Every headless / CI / agent run was blocked from codex image
gen even when fully authed.
Gate on the durable credentials file ($CODEX_HOME/auth.json) instead of
the TTY/stderr-only human text. Token validity is still proven by the
exec, which fails cleanly on a stale login. The stdout `features list`
capability check is unchanged.
Verified: reproduced the false 'not logged in' block, then after the fix
generated end-to-end via `resolve -t image --provider codex` (valid
1254x1254 PNG, source=generated, provider=codex.image_gen).
* fix(media-use): bug-bash fixes — id race, provider/reuse/adopt guards
From the bug-bash against main:
- MU-23 (HIGH): concurrent resolves raced on nextId (read-max-then-append,
non-atomic), so parallel agents got duplicate ids and clobbered each
other's files. Add allocateId(): a coarse per-project lock (.media/.lock,
15s stale-steal) around id allocation that scans the manifest AND the
type dir for reserved ids, then O_EXCL-creates a placeholder file so the
slow download between allocate and append can't collide. 5 parallel
resolves now yield 5 distinct ids + files.
- X4: --reuse imported across a type mismatch (bgm asset under images/).
Apply typesMatch on the --reuse path; reject mismatches (icon<->image
still interchangeable).
- X5: --provider silently overrode --local-only and made a network call.
--local-only is now a hard guard: network providers are skipped even
under a forced provider; the miss message explains the conflict.
- BUG-2: --provider ignored the exact-cache floor and could hand back an
asset from a different provider. A forced --provider now bypasses all
reuse rungs (regenerate with THIS provider); the unforced floor is intact.
- MU-26/X6: 0-byte assets accepted. --adopt skips 0-byte files (loud); ingest
refuses a 0-byte local file (freezeUrl already rejects empty responses).
- BUG-4: unknown/unavailable --provider now errors with the available list
instead of a generic 'no provider could resolve' (typo != catalog miss).
- BUG-5: --reuse "" gave the wrong 'type and intent required' error; it now
routes to a clear empty-sha message.
- BUG-3: voice duration leaked an unrounded float into index.md; round all
durations to 0.1s centrally at record build (matches probe).
- Nits: whitespace-only --intent is rejected; nudge grammar (exists/exist).
Tests: allocateId reservation + registry local-only-wins added; full
media-use suite green. All fixes verified e2e.
* fix(cli): reject unknown flags instead of silently ignoring them
citty is permissive: an unrecognized flag was dropped, not rejected — so
`render . --out x` (the flag is --output/-o) silently ignored --out and
rendered to the default renders/<name>.mp4 path. A mistyped flag read as a
render/catalog miss.
Add assertKnownFlags(): validate every dash-prefixed token against the
command's declared args + aliases + the global set (help/version/json)
before the command runs, in the shared trackCommandFailures run-wrapper so
every leaf command is covered. Handles --flag=value, --no-<bool> negation,
camelCase<->kebab arg names, and combined shorts; stops at --; positionals
and flag values pass through.
Verified: `render . --out x` -> 'Error: Unknown flag: --out'; --output/-o/
--json/--help still accepted. Unit tests added.
* docs(skills): install with --full-depth so agents get current main
The documented `npx skills add heygen-com/hyperframes` fetched the
skills.sh registry blob, which lags GitHub main by hours — so users
following the docs got a stale skill (e.g. media-use v1: no --candidates,
voice stubbed). The CLI's own `hyperframes skills` command already forces
a full clone via --full-depth to bypass this; the docs didn't pass it.
Add --full-depth to every documented install command (README, CLAUDE.md,
docs/guides/skills.mdx) with a one-line note on the lag. Addresses the
user-facing half of the publish/registry lag (#2034).
* chore(media-use): collapse resolve.mjs import to satisfy oxfmt --check
* fix(cli): extract longFlagName to keep flag validator under complexity gate
Also regenerate skills-manifest.json (resolve.mjs formatting change re-hashed
the media-use skill). Fixes the Fallow audit + skills-manifest-in-sync CI gates.
The ElevenLabs provider spawns a Python helper that writes straight to wavAbs
via a bare open(), which (unlike heygen/kokoro) never creates the parent dir —
so on a fresh project the save throws ENOENT and the line is silently dropped
as 'TTS failed - omitted'. mkdir -p the dir first, guarded so a mkdir failure
(EACCES/EROFS) returns { ok:false } like the rest of the branch rather than
throwing. (Migrated from #1960, whose skills/hyperframes-media path was retired
into skills/media-use; the bug moved with it.)
Reuse now hands the semantic judgment to the coding agent instead of a
string heuristic, while keeping the deterministic normalize-exact match as
an automatic dedup floor. No LLM/embedding call enters resolve; it stays
offline-capable.
- lib/match.mjs: shared matchTokens + typesMatch (extracted from adopt.mjs
and resolve.mjs so the icon<->image equivalence and token rules can't
drift between the do-path and the look-path); adds tokenOverlap ranker.
- resolve --candidates: side-effect-free listing of reusable assets across
the project manifest AND the global ~/.media cache, ranked by lexical
overlap, capped per scope, --json or human table. Never hard-filters on
zero overlap (that would pre-empt the agent's judgment); the agent decides.
- resolve --reuse <sha>: import a specific global-cache asset by content
sha/prefix (from --candidates) into the project via importFromCache,
marked source=reused-explicit / provenance.reused_by=agent.
- Adherence nudge: on a resolve that misses the exact floor and is about to
fetch, print a one-line stderr hint when similar cached assets exist,
pointing at --candidates. Offline, stderr (safe under --json), never
auto-reuses a fuzzy match.
- cache.mjs: export readGlobalManifest; add findGlobalBySha (prefix resolve
with ambiguity/miss handling).
- Telemetry: media_use_candidates event + reused-explicit source on
media_use_resolve (type/scope/counts only, no intent text or paths).
- SKILL.md: 'Reuse before you resolve' guidance + trust guardrail
(prefer-fresh-when-unsure, entity-exact for brand, cross-project bleed).
- Tests: lib/candidates.test.mjs (ranking, no-hard-filter, cap/truncation,
icon<->image, sha resolution, formatter); adopt.mjs refactor covered by
existing lib/adopt.test.mjs.
Full media-use suite green; verified e2e against the live catalog
(cross-project resolve->candidates->reuse; hint fires on miss).
Miguel Ángel