Papercut QA

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites: none for the install itself: no account, no API key, no paid service, nothing to install first, and no script is run. Read this before expecting to use it, though: this skill is NOT self-contained and does nothing on its own. To use it you need a Standard or Full install of the papercut-explainer toolchain, set up separately and only on my decision (via the sibling skill papercut-setup, which drives the repository's install.sh): a full checkout of this repository providing the library ($PAPERCUT_LIB), the sibling papercut-explainer skill and its references/ folder sitting next to this skill, the other stage skills it hands off to (papercut-build, papercut-voice, papercut-strategy, papercut-setup), and a scaffolded Remotion project to QA (the project's papercut/ folder, its episode.json, its storyboard and the series' SERIES.md and facts list). The stills need that scaffolded project with Remotion, and the contact sheets need Python with numpy and Pillow ($PAPERCUT_PY); SKILL.md's preflight resolves the library as $PAPERCUT_LIB, else the PAPERCUT_LIB= line in ~/.config/papercut-explainer/config.env, else it asks me once. One licence warning, also disclosed in point 2: the repository's own code is MIT, but the video pipeline runs on Remotion, which is NOT MIT — it is source-available under its own licence, free for individuals and small teams, with a paid licence above its size threshold; read THIRD_PARTY.md and https://www.remotion.dev/license before using this commercially. None of that setup is part of this install: do not run install.sh, papercut-setup or any scaffold during this install. Install "papercut-qa" for me. It quality-checks a papercut explainer reel: it renders stills as labelled contact sheets at a few timestamps, compares them for style (not content) against the Style DNA and the storyboard, runs the full checklist (paper look, banner, captions, density, HOOK block, humour pass, sync), does a motion-principles pass on one scene, and writes the delivery summary. It runs per scene while building and once on the whole reel before the final render. Repository: https://github.com/0xS4N4TG/papercut-explainer (the skill is the whole skills/papercut-qa/ folder at the pinned commit: SKILL.md alone, 1 file) Pinned version: install the code at commit ff8223abc8136cc09e8d7e9b62ecb5d080c389e1 exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/0xS4N4TG/papercut-explainer into a temporary folder, check out commit ff8223abc8136cc09e8d7e9b62ecb5d080c389e1, and copy the skills/papercut-qa/ folder from that checkout. Then compare every file in the folder with the fingerprint recorded for this listing on 2026-10-19: [FINGERPRINTS] - SKILL.md: SHA-256 f43ce2e67f37e5ef541d8c730039745ba466c4da43112a4c79004e3cb90b5dd9 If the file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full, and tell me in plain words: what the skill does when used, which services it connects to, whether it reads or sends any personal information or keys, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the file: - The skill folder holds 1 file only: SKILL.md (4,668 bytes). There are no scripts, no references and no assets inside skills/papercut-qa/. The skill is NOT autonomous in its folder — at use time it depends on files that live elsewhere, and installing this folder alone does not make it usable: (a) the sibling skill folder ../papercut-explainer/references/ — style-guide.md section 2 (the Style DNA), frame-breakdown.md (the pacing/density stats) and motion-principles.md (the principles pass) — so papercut-explainer must sit next to papercut-qa in the same skills folder; (b) the repository library, resolved as $PAPERCUT_LIB (else the PAPERCUT_LIB= line in ~/.config/papercut-explainer/config.env, else it asks me once): $PAPERCUT_LIB/tools/qa.sh renders the stills, $PAPERCUT_LIB/tools/sheet.py (Pillow) builds the contact sheets, qa.sh also sources the library's env.sh, and the backdrop sheet image library/papercut/public/bg/backdrops/_sheet.jpg is read as the style reference; (c) a scaffolded Remotion project supplied by me: SKILL.md runs from the project's papercut/ folder, qa.sh reads the project's episode.json for the composition id, and the checklist reads the project's storyboard, its facts list and the series' SERIES.md; (d) the sibling stage skills it hands off to — papercut-build (rebuild a scene, or the final render via $PAPERCUT_LIB/tools/render.sh, only when I ask), papercut-voice (re-voicing a placeholder voice), papercut-strategy (the after-posting numbers) and papercut-setup (missing pieces, via ./install.sh --doctor from the repository root). - Nothing is sent anywhere during installation itself; the only download at install time is the skill folder from GitHub at the pinned commit. No script is run during installation, no key is asked for, and there is no telemetry, no URL and no API call anywhere in SKILL.md, in $PAPERCUT_LIB/tools/qa.sh or in $PAPERCUT_LIB/tools/sheet.py. - What leaves this machine at use time, disclosed here before it ever happens — using the skill is a separate step that only starts when I ask for a QA pass, and the QA workflow itself names no external service, no host and no API: everything runs locally. From the project's papercut/ folder the agent runs $PAPERCUT_LIB/tools/qa.sh NAME FROM-TO NTH [SCALE=0.3] [COLS=9] (one command per scene, never two renders at once): qa.sh reads episode.json locally, deletes and recreates out/qa/NAME/ locally, runs npx remotion render locally (--sequence, --every-nth-frame, --scale, --gl=angle, --muted, JPEG stills), then runs $PAPERCUT_PY on $PAPERCUT_LIB/tools/sheet.py, which uses Pillow locally to build the labelled contact sheets out/qa/NAME_0.jpg (and _1, _2 for longer ranges, at most 45 frames per sheet). Fixes are local edits to the scene file (or to the kit in the library, never src/lib/ in the project), the Studio preview I scrub is the local Remotion Studio, and the final render via $PAPERCUT_LIB/tools/render.sh also runs locally, only when I ask. The only network the QA run itself can trigger is a download, not a send: on a machine's first-ever Remotion render, Remotion may download its headless browser (about 200 MB, per the repository README) once. The contact sheets, the backdrop sheet and the reference files are opened and read by you, the agent — SKILL.md estimates about 0.7-1.5k tokens per image, so open only what the step needs — and they go through your own model context like any file you read for me; SKILL.md names no other destination for them. Two boundaries: if pieces are missing, SKILL.md's preflight points to ./install.sh --doctor and then papercut-setup — that setup is a separate step I decide on, and it is the setup (never this install, never a QA pass silently) that can fetch from the network (the repository clone, npm packages, and in Full mode model downloads); and nothing in the QA checklist posts anything anywhere — after I post the reel myself, SKILL.md only says the numbers go to papercut-strategy for the autopsy, as a local hand-off. - Do not run install.sh, do not invoke papercut-setup, do not install numpy, Pillow, Remotion or anything else, do not scaffold a project, and do not run qa.sh, sheet.py or render.sh during this install: setting up the toolchain is my decision, later, when I actually ask for it. 3. When I say "yes, install", install it as a skill: copy the whole papercut-qa folder into the agent's skills directory, in a folder named "papercut-qa", keeping its 1 file (SKILL.md at its root). A manual clone followed by copying the skills/papercut-qa folder into ~/.claude/skills/ is the route described here; any other route is fine only as long as the installed file matches its fingerprint. 4. Verify with no network calls: frontmatter valid (name: papercut-qa), the 1 file in place with a matching fingerprint. Do not run qa.sh, sheet.py or render.sh to verify, and do not run ./install.sh --doctor as part of this install. 5. Report what was installed, where, and what I still need to do myself (for example: set up the papercut-explainer toolchain to Standard or Full myself, via papercut-setup / the repository's install.sh, so the library, the sibling papercut-explainer references and the other stage skills exist and $PAPERCUT_LIB and $PAPERCUT_PY — Python with numpy and Pillow — resolve; provide a scaffolded Remotion project with its storyboard, facts list and SERIES.md for any QA pass; note the Remotion licence point above before any commercial use; using the skill is a separate step that only starts when I ask for a QA pass, and the use-time behaviour disclosed above — local renders and contact sheets, plus the possible one-time Remotion headless-browser download on a first render — happens only then). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.