Files
hyperframes/scripts/test-skills-fresh.sh
WaterrrForever b9be0b2625 feat(skills,studio,media-use): the intent layer, review loop, and user memory — BRIEF.md, companion mode, recipes; /website-to-video folds into /product-launch-video (#2133)
* feat(studio,cli): per-frame board comments, self-refreshing storyboard, status-aware preview landing

Per-frame comment boxes on the storyboard board batch into
.hyperframes/frame-comments.json (a resubmit wins per frame; unconsumed
comments on other frames are kept). Submitted-but-unconsumed comments
stay visible — a toolbar banner plus a per-tile echo — until the agent
consumes the file; the banner also says what to do next (reply anything
in the agent chat).

The board keeps itself current: GET /projects/:id/signature exposes the
watcher-cached project signature, the storyboard payload carries the
signature it was derived from, and the view polls at 2s (hidden tabs
skipped, re-checked on visibility), refetching in place with no loading
flash. Posters bake the signature into their URL so tiles fill in as
sketches land and a poster that failed mid-write retries on the next
version; the empty state upgrades itself when STORYBOARD.md appears,
and its handoff prompt now points the agent at the review loop and uses
the parser's real status vocabulary (outline, not planned).

preview lands the browser on the storyboard view while the board is the
review surface — any frame built, or pure planning (srcs declared, none
on disk yet) — and on the timeline once the video is assembled.

* feat(skills): the review loop — plan, sketch, build as one shared process

hyperframes-core/references/review-loop.md is the single source for the
three-pass collaborative review: the plan proposed on a live board
(§ 1), wireframe sketches marked built with one layout question (§ 2 —
real words on plain blocks, run no CLI; a confirmed board is itself a
valid deliverable when the user asked for a storyboard, not a video),
the build dressing confirmed layouts (§ 3, worker or inline), and the
final look (§ 4). Autonomous runs skip every gate and keep one question
before render.

The three narrative workflows' Steps 3/4/6 collapse to references plus
their sketch stand-ins (captured-asset blocks for product-launch-video,
plain code panels for pr-to-video); the confirmed-sketch handoff stays
in each frame-worker prompt. general-video plans on a board for
multi-scene narrative pieces in collaborative mode — its sketch pass is
layout-before-animation with the user watching. The router treats
"I want a storyboard" as a process request rather than a route, and
closes exploratory intake by recommending a route plus how the run will
review.

The supporting contracts land next door: the comments channel (silent
submit, one reply picks it up, check the file before the words) in
brief-contract § 1; the sidecar schema and the built status rung in
storyboard-format; the mode question asked first and alone in the three
workflows' Step 0.

* feat(media-use): user memory — remembered preferences and frozen recipes

Two tiers of memory on media-use's existing two-tier storage split.

Preferences (lightweight): confirmed brief answers — destination, aspect,
language, mode, voice, style preset — recorded to the project's
.media/preferences.json (committed, the team inherits it) and promoted
to the personal ~/.media/preferences.json once the same value is
confirmed in two different projects (a sightings ledger accumulates the
cross-project evidence user-side, since project files can't see each
other). prefs.mjs get/record; merge reads project-over-user; a changed
value restarts its provenance.

Recipes (heavyweight): one approved run frozen as a named, versioned
bundle — frame.md, the storyboard skeleton (structure kept: durations,
transitions, srcs, Video direction; statuses reset to outline; content
blanked to per-frame fill-ins naming the beat's role), and the confirmed
brief values. Named folders, not content hashes: re-freezing bumps
version and archives <name>@v<N>; a freeze is already confirmed, so it
promotes to the user tier immediately. recipe.mjs freeze/list/use, plus
resolve --type recipe --entity <name> delegating like grade/lut.

16 new node --test cases; the media-use lib suite is 168/168.

* feat(skills): wire user memory into the brief and the review loop

brief-contract § 2 gains Remembered defaults: read the merged
preferences before Round 2 and let a remembered value become the
recommended option with a receipt naming its source project. Memory
changes the default, never the question — every ask-marked field still
gets asked, and what the request says this time beats what was picked
last time. Record only what the user actually confirmed (a defaulted
voice nobody chose is not an answer; a "go" that accepts the
recommended defaults is). The first record announces itself once;
after that the receipts carry the reminder. In autonomous mode a
remembered value becomes the decided value, receipt included.

The three narrative workflows read the remembered defaults before
Round 2, record the confirmed answers at the Step 0 gate, record the
chosen preset at the Step 2 gate (pr-to-video excepted — its preset is
fixed), and fall back to the remembered voice when the request names
none. general-video's discovery reads the same defaults.

Recipes wire in at both ends: Step 0 checks for a matching recipe
before the mode question — one question, plural-aware, and adopting
one fills the brief, skips the design step, and drafts the storyboard
from the frozen skeleton while every review gate still runs. The
review loop's final look (§ 4) offers the freeze once after approval,
and the confirmation teaches the recall phrase — the name is something
the system reminds the user of, never something they must remember.
The router recognizes a named recipe or "like last time" as a route.

* docs(skills): the sketch pass names check, not the deprecated validate

* feat(skills): intent-layer references — process, route briefs, capability menu, BRIEF.md format

* feat(media-use): brief skeleton as the recipe's fourth artifact; flow/storyboard preference keys

* feat(skills): the intent layer conducts every brief — workflows execute BRIEF.md

* feat(skills): retire the mode preference key; sync catalog surfaces for intent layer

* refactor(skills): dedupe router vs intent-layer guidance — one owner per rule

* feat(skills): the design ask — own spec, pick by eye from showcases, or defer

* docs(skills): the design ask says the honest line on capture routes

* feat(skills): product-launch-video absorbs website-to-video as the tour angle

* refactor(skills): keep product-launch-video pristine — a tour is brief intent, not a pipeline branch

* feat(skills): production loop + genre lenses; general-video goes freeform (route yours, laws hold)

* refactor(skills): /hyperframes is the front door - route tables and scope lists leave the workflows

* docs(skills): review-loop pass across skill catalog

* fix(cli): pass project dir to openStudioBrowser in background-server path

* feat(skills): add pitch-round reference - verbalized sampling concept gate

* feat(skills): wire pitch round into intent layer - completeness triage + route eligibility

* feat(skills): editorial capability recommendations, handoff disciplines, menu-probe split

* feat(skills): pitches carry their machinery; source-only-formed requests pitch the telling

* feat(skills): companion goes director - ceiling treatment plus blueprint/rule citation discipline

* fix(scripts): sandbox npx-leak guard - private npm global prefix keeps npx on the branch CLI

* chore(skills): resync manifest hash after formatter pass reflowed general-video tables

* fix(skills): recipe freeze reads workflow from BRIEF.md; style_preset records require workflow scope

Two holes found by a live companion-run freeze: the agent-supplied --workflow
contradicted the run's actual workflow (recipe.json said faceless-explainer,
brief-skeleton said general-video), and the style_preset lookup missed because
the preference had been recorded under the bare key.

- freezeRecipe resolves the workflow from BRIEF.md frontmatter; the flag is a
  fallback for briefless projects and a contradicting flag is ignored (noted).
- recordPreference refuses a bare style_preset — the scoped key is the only
  writable shape; freeze tolerates legacy bare records via read fallback.
- review-loop § 4 / media-use SKILL / brief-format wording follow the machinery.
2026-07-15 21:19:14 +08:00

303 lines
14 KiB
Bash
Executable File

#!/usr/bin/env bash
# test-skills-fresh.sh
#
# Generic sandbox for the `test/skills-fresh` branch — fully simulates a real
# user's `npx` install, with BOTH channels coming from the working tree:
# • skills → installed from skills/ via `npx skills add <repo>` (exactly what
# `npx skills add heygen-com/hyperframes` does for a real user)
# • CLI → wired via a `file:` dep so `npx hyperframes` resolves to the LOCAL
# build, which carries this branch's packages/cli/src/capture changes.
# It adds NO CLAUDE.md / AGENTS.md — it mirrors the plain install, nothing more.
# You then launch your agent in the sandbox and type whatever request you want.
#
# Agents: works for Claude Code (default) and Codex. `--agent` is passed straight
# to `skills add`, so the skills land in that agent's project dir:
# • claude-code → .claude/skills/ (launch: claude --dangerously-skip-permissions)
# • codex → .agents/skills/ (launch: codex --dangerously-bypass-approvals-and-sandbox)
# ← both launch fully auto (no approval prompts); codex stays
# project-local and does NOT touch your global ~/.codex/skills.
#
# Why a sandbox (and not `npx skills add heygen-com/hyperframes#test/skills-fresh`):
# `skills add` only copies skills/. The capture tool you changed lives in
# packages/cli (the @hyperframes/cli package), so an online skills-only install
# would pull this branch's skills but the PUBLISHED CLI's old capture. This
# script builds + file:-links the local CLI so capture comes from the branch too.
#
# Usage:
# bash scripts/test-skills-fresh.sh # Claude Code (default)
# bash scripts/test-skills-fresh.sh --agent codex # Codex
# bash scripts/test-skills-fresh.sh --rebuild # force a CLI rebuild
# bash scripts/test-skills-fresh.sh --no-build # skip the build step
# bash scripts/test-skills-fresh.sh -h # help
#
# What it does:
# 1. Verifies prerequisites (bun, npm, the chosen agent, optionally Chrome).
# 2. Builds the local CLI if dist/cli.js is missing OR any packages/cli source
# is newer than the built bundle (so your capture edits are never tested stale).
# 3. Creates a fresh WORKSPACE ROOT under /tmp/skills-fresh-<timestamp>/ with a
# package.json (`file:` CLI dep). It does NOT init a hyperframes project here
# — the video workflows run `npx hyperframes init` inside their own subdirs.
# 4. Runs npm install (file: dep), then installs the CLI into a sandbox-private
# npm global prefix ($TEST_DIR/.npm-global). npx checks that prefix before its
# ~/.npm/_npx cache, so `npx hyperframes` resolves to the LOCAL build from any
# cwd — agents run it from scratchpads / home after shell resets, where the
# file: dep is invisible and npx would silently use the published CLI.
# 5. Installs the full skills tree from the LOCAL repo via `npx skills add
# --agent <agent>`, then prunes the internal _meta/ authoring skills so the
# installed set matches what an end user gets.
# 6. Verifies the router + 10 workflows + 6 domain skills landed.
# 7. Prints the command to start the agent + example prompts to try.
#
# Iterate after editing:
# • skills → re-run this script (fresh dir), or in the existing test dir:
# rm -rf <skills-dir>/<name> && npx --yes skills add <repo> \
# --skill <name> --agent <agent> --yes
# • capture / CLI → just re-run this script; step 2's staleness check rebuilds.
set -uo pipefail
# --------- defaults ---------
EXPECTED_BRANCH="test/skills-fresh"
AGENT="claude-code"
# --------- arg parse ---------
BUILD_MODE="auto" # auto | force | skip
while [[ $# -gt 0 ]]; do
case "$1" in
-h|--help)
sed -n '2,40p' "$0" | sed 's/^# \{0,1\}//'
exit 0
;;
--agent) AGENT="${2:-}"; shift 2 ;;
--rebuild) BUILD_MODE="force"; shift ;;
--no-build) BUILD_MODE="skip"; shift ;;
*)
echo "Unknown arg: $1" >&2
echo "Run with --help for usage." >&2
exit 1
;;
esac
done
if [[ -z "$AGENT" ]]; then
echo "--agent needs a value (e.g. claude-code, codex)" >&2
exit 1
fi
# Map the agent to its project skills dir + launch binary.
case "$AGENT" in
claude-code) SKILLS_DIR=".claude/skills"; AGENT_BIN="claude"; LAUNCH="claude --dangerously-skip-permissions" ;;
codex) SKILLS_DIR=".agents/skills"; AGENT_BIN="codex"; LAUNCH="codex --dangerously-bypass-approvals-and-sandbox" ;;
*) SKILLS_DIR=".agents/skills"; AGENT_BIN="$AGENT"; LAUNCH="$AGENT" ;;
esac
# --------- self-locate the hyperframes repo ---------
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
HF_REPO="$(cd "$SCRIPT_DIR/.." && pwd)"
HF_CLI_PKG="$HF_REPO/packages/cli"
HF_CLI_BIN="$HF_CLI_PKG/dist/cli.js"
# --------- pretty output helpers ---------
say() { printf "\033[1;36m→ %s\033[0m\n" "$*"; }
ok() { printf " \033[0;32m✓\033[0m %s\n" "$*"; }
warn() { printf " \033[0;33m! %s\033[0m\n" "$*"; }
fail() { printf " \033[0;31m✗ %s\033[0m\n" "$*"; exit 1; }
# --------- step 1: prerequisites ---------
say "Checking prerequisites (agent: $AGENT)..."
command -v bun >/dev/null 2>&1 || fail "bun not installed. Install: curl -fsSL https://bun.sh/install | bash"
command -v npm >/dev/null 2>&1 || fail "npm not installed (need Node.js — install Node 22+)."
ok "bun: $(bun --version)"
ok "node: $(node --version)"
ok "npm: $(npm --version)"
if command -v "$AGENT_BIN" >/dev/null 2>&1; then
ok "$AGENT_BIN on PATH"
else
warn "$AGENT_BIN not on PATH — install the $AGENT CLI before running the test."
fi
CHROME_MAC="/Applications/Google Chrome.app/Contents/MacOS/Google Chrome"
CHROME_LINUX="/usr/bin/chromium"
if [[ -x "$CHROME_MAC" ]] || [[ -x "$CHROME_LINUX" ]]; then
ok "Chrome / Chromium found (capture / web-extraction needs headless Chrome)"
else
warn "No Chrome at $CHROME_MAC or $CHROME_LINUX — capture-based workflows will fail without it"
fi
CURRENT_BRANCH="$(cd "$HF_REPO" && git rev-parse --abbrev-ref HEAD 2>/dev/null || echo unknown)"
if [[ "$CURRENT_BRANCH" != "$EXPECTED_BRANCH" ]]; then
warn "Repo is on '$CURRENT_BRANCH', not '$EXPECTED_BRANCH' — you'll be testing whatever is checked out."
else
ok "repo on branch: $CURRENT_BRANCH"
fi
# --------- step 2: build local CLI (staleness-aware) ---------
say "Checking local CLI build..."
needs_build() {
[[ ! -f "$HF_CLI_BIN" ]] && return 0
# any CLI source (incl. capture/) newer than the built bundle → rebuild,
# so your working-tree edits are never silently tested against a stale dist.
local newer
newer="$(find "$HF_CLI_PKG/src" "$HF_CLI_PKG/scripts" -type f -newer "$HF_CLI_BIN" 2>/dev/null | head -1)"
[[ -n "$newer" ]] && return 0
return 1
}
DO_BUILD=0
case "$BUILD_MODE" in
force) DO_BUILD=1; warn "--rebuild: forcing a fresh CLI build" ;;
skip) warn "--no-build: skipping build; using existing dist (may be stale!)" ;;
auto) if needs_build; then
DO_BUILD=1
[[ -f "$HF_CLI_BIN" ]] && warn "CLI source is newer than dist — rebuilding to pick up your capture/CLI edits" \
|| warn "CLI not built — building (~1-2 min)..."
fi ;;
esac
if [[ "$DO_BUILD" == "1" ]]; then
(cd "$HF_REPO" && bun install && bun run build) || fail "CLI build failed."
[[ -f "$HF_CLI_BIN" ]] || fail "Build completed but $HF_CLI_BIN still missing."
fi
ok "local CLI: $(node "$HF_CLI_BIN" --version 2>/dev/null || echo unknown)"
# --------- step 3: scaffold a fresh test project ---------
TEST_PARENT="${TEST_PARENT:-/tmp}"
TEST_NAME="skills-fresh-$(date +%H%M%S)"
TEST_DIR="$TEST_PARENT/$TEST_NAME"
say "Creating test project at $TEST_DIR ..."
mkdir -p "$TEST_PARENT"
cd "$TEST_PARENT"
[[ -e "$TEST_NAME" ]] && fail "$TEST_DIR already exists. Wait 1s and re-run."
# WORKSPACE ROOT, not a hyperframes project: the video workflows run
# `npx hyperframes init` inside their own subdirs, so a project at the root would
# make a skill find a stray composition here. We only need a package.json with the
# `file:` CLI dep so `npx hyperframes` (and the skills' init/render calls from
# subdirs) resolve to the local build.
mkdir -p "$TEST_NAME"
cd "$TEST_NAME"
cat > package.json <<JSON
{
"name": "$TEST_NAME",
"private": true,
"type": "module",
"dependencies": {
"hyperframes": "file:$HF_CLI_PKG"
}
}
JSON
ok "package.json points hyperframes → file:$HF_CLI_PKG"
# --------- step 4: npm install (NOT bun) ---------
# MUST be npm: bun follows the cli pkg's `workspace:*` devDependencies and fails.
# npm only resolves the file: package's `dependencies`.
say "Running npm install (must be npm here, not bun)..."
npm install --no-audit --no-fund --silent || fail "npm install failed."
[[ -x "node_modules/.bin/hyperframes" ]] || fail "node_modules/.bin/hyperframes missing after install."
ok "node_modules/.bin/hyperframes → local CLI"
# --------- step 4.5: sandbox-private npm global prefix (npx-leak guard) ---------
# The file: dep only covers `npx hyperframes` run somewhere UNDER $TEST_DIR — npx
# walks up from cwd looking for node_modules/.bin. Agents routinely run it from
# OUTSIDE the tree (session scratchpads, home after a shell cwd reset); there the
# walk finds nothing, npx falls back to ~/.npm/_npx, and the run silently tests
# the PUBLISHED CLI instead of the branch (same version string, so nothing warns).
# npx checks the npm global prefix BEFORE that cache, so installing the local CLI
# into a sandbox-private prefix — exported at launch — closes the leak from any cwd.
say "Installing local CLI into sandbox npm global prefix (npx-leak guard)..."
NPM_GLOBAL_PREFIX="$TEST_DIR/.npm-global"
npm install -g "file:$HF_CLI_PKG" --prefix "$NPM_GLOBAL_PREFIX" --no-audit --no-fund --silent \
|| fail "global-prefix install failed."
[[ -x "$NPM_GLOBAL_PREFIX/bin/hyperframes" ]] || fail "$NPM_GLOBAL_PREFIX/bin/hyperframes missing after install."
ok "npx hyperframes now resolves to the local build from ANY directory (launch with the env below)"
# --------- step 5: install skills from the local repo, then prune _meta ---------
say "Installing skills from the local repo (--agent $AGENT) ..."
npx --yes skills add "$HF_REPO" --skill '*' --agent "$AGENT" --yes \
|| fail "skills add failed."
# Resolve where they actually landed (claude-code → .claude/skills,
# codex/others → .agents/skills); fall back to whichever dir got populated.
if [[ ! -d "$SKILLS_DIR" ]]; then
for d in .claude/skills .agents/skills .cursor/skills; do
[[ -d "$d" ]] && SKILLS_DIR="$d" && break
done
fi
[[ -d "$SKILLS_DIR" ]] || fail "No skills dir found after install (looked for .claude/skills, .agents/skills, .cursor/skills)."
ok "skills installed under $SKILLS_DIR/"
# skills add walks skills/_meta/ too — those are internal authoring skills, not
# part of the end-user set. Prune them so the sandbox matches a real install.
if [[ -d "$HF_REPO/skills/_meta" ]]; then
for meta in "$HF_REPO/skills/_meta"/*/; do
[[ -d "$meta" ]] || continue
name="$(basename "$meta")"
if [[ -d "$SKILLS_DIR/$name" ]]; then
rm -rf "$SKILLS_DIR/$name"
ok "pruned internal meta-skill: $name"
fi
done
fi
# --------- step 6: verify the skills landed ---------
say "Verifying skill installation..."
ROUTER="hyperframes"
WORKFLOWS=(product-launch-video faceless-explainer embedded-captions \
talking-head-recut pr-to-video motion-graphics general-video \
remotion-to-hyperframes slideshow)
DOMAIN=(hyperframes-core hyperframes-creative hyperframes-animation hyperframes-cli media-use hyperframes-registry)
MISSING=()
check_skill() { if [[ -d "$SKILLS_DIR/$1" ]]; then ok "$SKILLS_DIR/$1/"; else MISSING+=("$1"); fi; }
check_skill "$ROUTER"
for s in "${WORKFLOWS[@]}"; do check_skill "$s"; done
for s in "${DOMAIN[@]}"; do check_skill "$s"; done
if [[ ${#MISSING[@]} -gt 0 ]]; then
warn "Missing skill(s): ${MISSING[*]}"
warn "Check skills/ in the repo and re-run — routing / dispatch will break without these."
fi
INSTALLED_COUNT=$(find "$SKILLS_DIR" -maxdepth 1 -mindepth 1 -type d 2>/dev/null | wc -l | tr -d ' ')
ok "$INSTALLED_COUNT skill(s) installed under $SKILLS_DIR/"
# --------- step 7: print next steps ---------
echo ""
printf "\033[1;32m========================================================\033[0m\n"
printf "\033[1;32m Sandbox ready — branch skills + branch CLI (capture).\033[0m\n"
printf "\033[1;32m========================================================\033[0m\n"
echo ""
echo "Project: $TEST_DIR"
echo "Agent: $AGENT (skills in $SKILLS_DIR/)"
echo "CLI: file:$HF_CLI_PKG (local build — includes your capture changes)"
echo "Branch: $CURRENT_BRANCH"
echo ""
echo "To start, run (the env line is the npx-leak guard — without it, npx run outside"
echo "the sandbox tree silently falls back to the PUBLISHED CLI in ~/.npm/_npx):"
echo ""
printf " \033[1;37mcd %s\033[0m\n" "$TEST_DIR"
printf " \033[1;37mnpm_config_prefix=\"%s\" PATH=\"%s/bin:\$PATH\" %s\033[0m\n" "$NPM_GLOBAL_PREFIX" "$NPM_GLOBAL_PREFIX" "$LAUNCH"
echo ""
echo "Then type any request you want to test — the agent routes it to a workflow. e.g.:"
echo " • \"make a product launch video for https://your-site.com/\" → product-launch-video (exercises capture)"
echo " • \"explain how transformers work as a faceless explainer video\" → faceless-explainer"
echo " • \"make a video from this PR: owner/repo#123\" → pr-to-video"
echo " • \"add lower-thirds / overlay cards to ./clip.mp4\" → talking-head-recut"
echo " • \"add captions/subtitles to ./clip.mp4\" → embedded-captions"
echo " • \"turn https://your-site.com/ into a site tour video\" → product-launch-video"
echo " • \"a logo reveal / title card / data montage\" → general-video"
echo ""