Version-controlled reusable skills for Permaculture Works software, operations, and design workflows.
skills/web-security-quality-audit/ — repeatable evidence-based web security, privacy, reliability, accessibility, performance, maintainability, and release-readiness auditing.
The skill is reusable. Project-specific routes, URLs, provider settings, commands, and risk gates stay in the target repository.
1. Read target AGENTS.md + audit.config.json.
2. Validate the target ontology if present.
3. Follow skills/web-security-quality-audit/SKILL.md.
4. Use current source/CI/live evidence.
5. Save a dated audit for future delta comparison.
The library models its own active architecture in:
ontology/SELF_MODEL.json— compact agent-facing index.ontology/SELF_MODEL.graphml— visualization/tooling projection.ontology/semantics.json— small hand-maintained semantic overlay.
The source tree remains canonical. Generated models contain no timestamp or commit SHA, so wording-only edits do not create churn. The topology fingerprint changes only when modeled structure or semantic relationships change.
node scripts/build-ontology.mjs --write
node scripts/validate-library.mjsCI reruns validation on pull requests, main, and weekly. A stale ontology fails validation.
- Keep
SKILL.mdconcise; put detail inreferences/, deterministic helpers inscripts/, and output scaffolding inassets/. - Keep reusable logic here and project-specific facts with the project.
- Use reviewed commits/tags for installed copies.
- Never embed credentials, secrets, or client data.
- Missing evidence is Unverified, never an assumed pass.
- Prefer smaller composable skills over one growing monolith.
See AGENTS.md and docs/skills-architecture.md.
The root AGENT.md and older site-model materials are retained as source material from the Permaculture Design Agent Core template. They are not repository-wide instructions for this skill library and are intentionally represented as a single legacy node in the self-ontology until they are decomposed into reusable skills.