Security audit
Security checks for vulnerabilities and agentic risk
ClawScan marks this skill suspicious because it found an unauthenticated rendering server with a file exposure risk, alongside other concerns about external services and installation; some files were only partially analyzed.
Consider avoiding installation until the rendering-server exposure is reviewed and the affected checks are addressed. If you proceed, use a pinned installer version, and avoid sending confidential scripts to external narration services unless you understand and approve that disclosure. The scan covered only 16 of 23 components fully, so its evidence is incomplete.
scripts/capture.tsUnauthenticated rendering server permits sibling-directory traversal
Attackers with network access during rendering can retrieve known readable files from sibling directories sharing the project's name prefix. The listener also makes project files available without authentication. This is not unrestricted filesystem traversal: unrelated path prefixes are rejected.
Referenced artifact was not completely inspected
with headless Chrome (scripts/capture.ts) or HyperFrames, and QA scripts that check label overlap, phone
Referenced artifact was not completely inspected
bun scripts/capture.ts . --stills 2.6,8.4,20.5 # look at qa/stills/*.png first
Referenced artifact was not completely inspected
bun scripts/capture.ts . --out renders/master.mp4 --labels qa/labels.json
Referenced artifact was not completely inspected
| `scripts/capture.ts` | render MP4 / stills / label boxes / determinism check / `--eval` from any page that publishes `window.__film`; system Chrome over DevTo
Referenced artifact was not completely inspected
bun scripts/check_labels.ts qa/labels.json # overlap, safe area, phone size, reading time
Referenced artifact was not completely inspected
- `check_labels.ts`: no overlapping text, 64 px side margins, names ≥ 46 px (9 px on a phone), every
Referenced artifact was not completely inspected
| `scripts/check_labels.ts` | overlap, safe area, anchor covered, phone size, reading time on the real render |
Referenced artifact was not completely inspected
bun scripts/check_facts.ts factsheet.md qa/labels.json # every word on screen is in the sheet
Referenced artifact was not completely inspected
- `check_facts.ts`: on-screen text equals the sheet, every claim sourced, critic verdict PASS;
Referenced artifact was not completely inspected
| `scripts/check_facts.ts` | on-screen text vs the fact sheet, sources, critic verdict, approval |
Referenced artifact was not completely inspected
bun scripts/sheet.ts renders/master.mp4 qa/sheet.png --n 24 --cols 6 --tile 400
Referenced artifact was not completely inspected
bun scripts/sheet.ts renders/master.mp4 qa/phone.png --phone # 390 px tiles = a phone feed
Referenced artifact was not completely inspected
| `scripts/sheet.ts` | contact sheets from a video or a stills folder, `--phone` 390 px tiles, `--at` times |
Referenced artifact was not completely inspected
Then replace `factsheet.md`, `story.json` and `cell.js` with your subject; `lib/` is the engine.
Referenced artifact was not completely inspected
| `cell.js` | the subject: geometry, placement, anchors, `update(t)` |
The documented npx command can download and execute an unpinned version of the skills installer with the user's privileges. Although this is an installer rather than an MCP server, a compromised or unexpected future package release could execute harmful code; nothing here indicates malicious author intent.
Without declared permissions the skill's intent is opaque and cannot be validated.
The optional narration workflow directs script content to varg and ElevenLabs without a privacy checkpoint for confidential material. Naming the providers and requiring cost approval reduce surprise, but do not ensure that the user understands or authorizes disclosure of sensitive script content.
Suspicious Unicode normalization or mixed-script content
The instructions explicitly send the entire narration script to external speech services, potentially exposing unpublished research or other confidential content. This is task-related processing rather than covert exfiltration, but approval of the quoted price is not necessarily informed approval of sensitive-data disclosure.
Suspicious Unicode normalization or mixed-script content
Any row whose claim field contains 'no claim' bypasses the missing-source check, even when it is not a title or actually asserts a factual claim. An untrusted or incorrectly generated fact sheet can therefore make unsupported statements pass this QA gate, although this affects content integrity rather than enabling code execution.