update-documentation — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited update-documentation (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
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.
Run inline — do NOT spawn a subagent. Your live context of what changed is the primary input. Delegating forces a lossy re-briefing and causes gaps.
Each row is a file and what it owns. Use this to reason about which files are affected by a change.
| File | Owns |
|---|---|
docs/SPEC.md | Authoritative algorithm steps, flags, output formats, error conditions for every command |
docs/UX.md | Prompt flows, decision matrices, flag interaction examples, sample terminal output |
docs/ARCH.md | Module list with one-liner APIs, data flow, tech decisions |
docs/ROADMAP.md | Phase status — mark phases complete here |
docs/SECURITY.md | Threat model, scan pipeline, chunking strategy |
packages/core/src/config.ts → DEFAULT_CONFIG_TEMPLATE | TOML shown to users on first run — must match schema |
website/guide/getting-started.md | First-install flow, scope, basic agent symlinks intro |
website/guide/installing-skills.md | Install behavior end-to-end: sources, scope, agent symlinks (incl. "skipped when" lists), conflict handling, multi-skill, security, flags, remove, link |
website/guide/configuration.md | Config wizard UX, TOML keys by section, example blocks |
website/guide/security.md | What skilltap scans for, thresholds, semantic scan |
website/guide/taps.md | Tap add/remove/list/update, tap.json format |
website/guide/creating-skills.md | SKILL.md format, frontmatter, verify command |
website/reference/cli.md | Every command: synopsis, flags table, behavior notes, examples |
website/reference/config-options.md | Every config key: type, default, description, full TOML example |
website/public/llms-full.txt | Generated — never edit directly. Regenerate with cd website && bun scripts/gen-llms-txt.ts after any guide or reference page changes. |
website/public/llms.txt | Manually maintained page-level index (llms.txt standard). Links to guide and reference pages — not individual commands. Update only when a new page is added/removed, the site description changes, or the Optional section changes. |
~/.claude/projects/…/memory/MEMORY.md | Phase key notes, gotchas, stable patterns for future sessions |
1. Grep before reading. Search the changed feature's name across docs/ and website/ to find gaps and stale text. Read only the relevant section (offset+limit), not the whole file.
2. Reason from the map. For each piece of changed behavior, ask: which files in the map own that area? Check those. Don't limit yourself to a fixed checklist — a change often touches multiple categories. Examples:
SPEC.md, UX.md, installing-skills.md, cli.mdconfig.ts template, SPEC.md, UX.md, configuration.md, config-options.mdSPEC.md, UX.md, cli.md, relevant guide page(s)UX.md decision matrix, guide page "skipped when" list, cli.md notes3. Guide pages own the narrative. website/guide/ pages are where users learn behavior — update prose, code block examples, and condition lists. website/reference/ pages need their tables kept accurate. Both matter.
*4. Regenerate `llms-full.txt` if any `website/guide/.md or website/reference/.md` changed.*
cd /home/nathan/dev/skilltap/website && bun scripts/gen-llms-txt.tsThis concatenates all pages in sidebar order into a single Markdown file for AI bulk ingestion. The script is the source of truth for page order — if a new page is added, add it to PAGES in website/scripts/gen-llms-txt.ts.
5. Update `llms.txt` only when page structure changes. website/public/llms.txt is a manually maintained index following the llms.txt standard. It lists pages — not individual commands or flags. Update it when:
Do not update it for: new commands on existing pages, new config keys, flag changes, or content edits within an existing page.
6. Update `MEMORY.md` for new stable patterns, gotchas, or completed phases. Keep it under 200 lines — be terse, put detail in a topic file if needed.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.