Add static skill browser UI at /docs

[yalbantov85]

Summary

• Adds a single-page, dependency-free browser for the 754-skill catalogue, deployable via GitHub Pages from /docs with no build step
• New tools/build-browser-data.py aggregates SKILL.md frontmatter + the MITRE ATT&CK Navigator layer into docs/data.json
• .github/workflows/update-index.yml extended to keep docs/data.json in sync alongside index.json on every push that touches skills/**
• UX features: filter by subdomain / tag / framework, full-text search across name/description/tags/technique IDs, URL hash state for shareable filtered views, keyboard navigation, active-filter chips, search-within-filter, pretty-formatted skill names with security acronym preservation (JWT, SOC, SIEM, ATT&CK, …), match highlighting

How to preview

python3 tools/build-browser-data.py
python3 -m http.server --directory docs 8000
# open http://localhost:8000

To deploy after merge: Settings → Pages → Source: Deploy from a branch → main / /docs. Pages auto-serves the static site at https://mukul975.github.io/Anthropic-Cybersecurity-Skills/.

Test plan

• python3 tools/build-browser-data.py runs from repo root and emits docs/data.json with 754 skills
• Open docs/index.html via a local HTTP server — search, subdomain, tag, and framework filters all narrow results
• URL hash updates on filter change; reloading the page restores the filtered view (#sub=cloud-security&q=lateral)
• ↑ / ↓ walks the result list; Enter opens the first result when nothing is active
• Active-filter chips appear above the results; clicking the × on a chip removes only that filter
• Existing validate-skills.yml workflow still passes on this branch
• update-index.yml workflow commits both index.json and docs/data.json on the next skills/** change

Notes for review

• All HTML insertion uses document.createElement + .textContent (no innerHTML). XSS-safe by construction even if a future PR adds adversarial frontmatter content.
• No new runtime dependencies; the data builder uses only the Python stdlib.
• docs/data.json is treated like index.json — CI-managed, do not hand-edit. The PR includes a generated copy so reviewers can preview without running the script.

🤖 Generated with Claude Code

[mukul975]

Tested locally — build runs clean, all 754 skills indexed, filters and search work correctly.

Wrote docs/data.json: 754 skills, 45 subdomains, 723.7 KB

Verified:

• XSS-safe throughout: all skill content rendered via textContent/createTextNode, zero innerHTML — and the comment in highlightInto() explicitly documents this
• Build script handles missing/malformed frontmatter gracefully (bad file skipped, not a crash)
• All files opened encoding="utf-8" explicitly
• CI regenerates docs/data.json on every skill push and commits it — no manual drift path
• No external CDN dependencies, no SKILL.md content touched

One known limitation worth a follow-up issue: the frontmatter parser doesn't handle multi-line folded scalars (description: >- spanning multiple lines). Currently skills/performing-nist-csf-maturity-assessment/SKILL.md renders with a blank description in the browser because of this. The skill still appears in search results — it's cosmetic only — but worth fixing the parser to handle >/>- blocks so it stays consistent with validate-skill.py.

Approving. Great addition to the repo.

[mukul975]

One thing before merge: both commits include Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>. Please squash to a single commit and remove that line — attribution should be yours only.