# Project Wiki

> Create, scan, update, sync, and maintain .project-wiki/ project knowledge bases, including requirements, document intake, and traceability. Consult and update an existing project wiki during coding tasks.

- Skill: `giodra96/project-wiki` (Agent Skill, multi-file: 7 files)
- Install (CLI): `npx skillmds@latest add giodra96/project-wiki`
- Raw SKILL.md: https://api.skillmd.com/api/skills/giodra96/project-wiki/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: giodra96 (https://skillmd.com/u/giodra96)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/giodra96/project-wiki

---


# Project Wiki

Maintain an agent-readable knowledge base in `.project-wiki/`. Load only the workflow and project context needed for the task.

Current project wiki schema version: `1.6.0` (canonical value: `./schema/project-wiki.yml`).

## Route The Task

Select one primary mode for an explicit wiki request and load only its workflow:

| Trigger | Load |
| --- | --- |
| `init` - Initialize an empty project wiki | [Init Workflow](./references/initialization-workflows.md#init-workflow) |
| `scan` - Scan an existing repository | [Scan Existing Project Workflow](./references/initialization-workflows.md#scan-existing-project-workflow) |
| `update` - Process notes, requirements, meetings, or durable analysis | [Update Workflow](./references/update-workflows.md#update-workflow) |
| `update` - Ingest PDF, DOCX, text, or Markdown sources | [Document Intake Workflow](./references/update-workflows.md#document-intake-workflow), then [Document Ingestion](./references/document-ingestion.md) |
| `sync` - Reconcile code changed outside the current agent task | [Sync Workflow](./references/synchronization-workflow.md#sync-workflow) |
| `maintain` - Validate, migrate, repair, or semantically lint a wiki | [Maintain Workflow](./references/maintenance-workflows.md#maintain-workflow) |
| Before implementing, modifying, debugging, refactoring, testing, documenting, or planning code when `.project-wiki/INDEX.md` exists | [Automatic Context Preflight](./references/automatic-workflows.md#automatic-context-preflight) |
| After agent-made source changes | [Automatic Post-Implementation Wiki Update](./references/automatic-workflows.md#automatic-post-implementation-wiki-update) |

Use [Document Templates](./assets/document-templates.md) when creating artifacts and [Requirement Authoring](./assets/requirements-change-templates.md#requirements-topic-scaling) for atomic REQ/NFR/CON records. Read [Wiki Structure](./references/wiki-structure.md) or [Schema Contract](./schema/project-wiki.yml) only for exact structure or contract values. Finalize edits with [Common Policies](./references/common-policies.md), including reconciliation and audit logging.

## Operating Contract

1. Read `.project-wiki/INDEX.md` first when it exists, follow its routes, and open only the smallest relevant set of files. Use `REGISTRY.yml` only when indexes do not resolve the task.
2. Run the automatic workflows around code work when the wiki exists; reserve `sync` for external or manual changes.
3. Follow the selected workflow completely. Prefer the supplied deterministic scripts over reproducing structural, validation, duplicate, extraction, or ledger logic in model reasoning.
4. Finalize every meaningful edit with the shared checklist and audit log; reconcile only affected records.

Use `scripts/wiki_scope.py` (`list`, `read`, `search`, `diff`) for repository source access during wiki work; helpers apply `.project-wiki/.wikiignore` automatically. Consult existing wiki history normally. Load [Repository Source Scope](./references/repository-scope.md) only for configuration or troubleshooting.

## Invariants

- Keep stable IDs, required frontmatter, relative wiki links, and the complete canonical structure. Preserve history and original requirements rather than silently replacing them.
- Put explicit business or product intent in `requirements/`; put implemented or observed behavior in `technical/`. Record unconfirmed product intent inferred from code as an open question.
- Reconcile existing open questions before creating new ones. Represent significant risks, contradictions, blocking gaps, or critical assumptions as alerts and resolve rather than delete them.
- Mark uncertain scan findings as `confidence: inferred` or `confidence: unknown`.
- Keep indexes as compact routing maps. File durable non-canonical synthesis only when significant and link it to evidence or related records.
- Treat document intake as immutable provenance. Never load external PDF or DOCX files directly into model context; use the ingestion and review-progress scripts and complete every chunk disposition.
- Write all wiki content and generated instruction blocks in English. Reply in the user's language unless requested otherwise.