Project Knowledge
This skill creates a small set of docs that explain a project in plain language.
Use it when:
- you start working with a new repo
- the repo has no clear documentation
- you keep repeating the same project context in chats
- you want Codex to understand the project faster on future tasks
Main benefit:
- less repeated explanation
- faster onboarding into any repo
- clearer handoff between planning, coding, and review
Output
Create or update:
docs/project_knowledge/project.mddocs/project_knowledge/architecture.mddocs/project_knowledge/patterns.mddocs/project_knowledge/deployment.mddocs/project_knowledge/ux-guidelines.mdonly if the project has meaningful UI
For file responsibilities and section guidance, read file-set.md.
Workflow
- Read the repository structure and the current docs state.
- If docs already exist, preserve useful content and remove stale or duplicate content.
- Read only the code and configs needed to establish:
- what the project does
- who it serves
- major components
- important dependencies
- how it runs and deploys
- Ask only for missing decisions that cannot be inferred safely.
- Write or update the project knowledge files in concise prose.
- Keep root-level instructions and README minimal. Durable operating knowledge belongs in
docs/project_knowledge/.
Rules
- Prefer facts tied to the current repo over generic framework explanations.
- Do not paste code blocks into these docs unless absolutely necessary.
- Link to source files when implementation details matter.
- Keep information in one place only. Cross-reference instead of duplicating.
- Record operational details that are hard to recover from code alone: env var names, service names, logs, deploy entry points, runtime assumptions.
Separation
Do not confuse these artifacts:
docs/project_knowledge/: stable project truthdocs/project_status_YYYY-MM-DD.md: point-in-time operator statuswork/<feature>/: temporary feature planning and execution docs