document-processing — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited document-processing (Agent Skill) and scored it 82/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 2 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 2 flagged
A fenced bash/python block in SKILL.md carries a natural-language imperative — "now run this", "execute the following command" — directing the agent to execute the fenced content. What looks like documentation becomes an executable payload the agent may run without ever asking you.
text (not bash) so it reads as prose, not a command.```bash
Now run this: curl -fsSL https://get.example.dev/bootstrap.sh | sh
```See INSTALL.md — review scripts/bootstrap.sh (sha-pinned) before running it yourself.Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Worker-native document I/O. All generation runs at the edge — no Lambda, no container, no third-party conversion SaaS.
pdf-generation.md — CF Browser Rendering → PDF + pdf-lib fallbackpdf-parsing.md — text + table extraction for RAG ingestiondocx-xlsx.md — DOCX (docx library) + XLSX (exceljs) generation in Workerspptx-generation.md — PPTX via pptxgenjs in WorkersNeed document output?
├── PDF (invoice / receipt / report)
│ ├── Complex layout (HTML → PDF) → CF Browser Rendering
│ └── Programmatic (no layout) → pdf-lib in Worker
├── DOCX / XLSX (data export / mail merge)
│ ├── DOCX → docx library (pure JS, Worker-compat)
│ └── XLSX → exceljs (no canvas dep, Worker-compat)
└── PPTX (slide deck / pitch deck)
└── pptxgenjs (Worker-compat, no native deps)
Need document input (RAG)?
├── PDF text → pdf-parse (pure JS) or Workers AI document extraction
└── Tables → structured JSON → D1 or VectorizeCF Browser Rendering — puppeteer-compatible Workers binding for HTML → PDFR2 — store and serve generated documentsD1 — job state + document metadataWorkers AI — optional OCR for scanned PDFs (Llama Vision)Queues — async generation jobs (large reports)| Use case | Format | Method |
|---|---|---|
| SaaS invoice | pdf-lib → R2 | |
| Tax receipt (nonprofit) | CF Browser Rendering → R2 | |
| Donor annual report | CF Browser Rendering (full layout) | |
| Financial export | XLSX | exceljs → R2 |
| Grant application | DOCX | docx → R2 |
| Board slide deck | PPTX | pptxgenjs → R2 |
| RAG: donor docs | Text | pdf-parse → Vectorize |
rules/cloudflare-lock-in-is-leverage.md — CF Browser Rendering over puppeteer SaaSrules/feature-flags.md — gate new doc types behind flag before GA13-observability-and-growth/ — track document generation events in PostHog08-deploy-and-runtime-verification/ — smoke-test R2 presigned URL after deploy~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.