mk:project-context

Project Context Generator

Creates docs/project-context.md — the agent "constitution" that every agent loads on activation.

To generate the full documentation suite (README, architecture, code standards, etc.), use mk:docs-init instead. mk:project-context generates only the agent constitution.

Purpose

Eliminate context drift between agents. Without a single authoritative file, agents infer conventions independently and make conflicting decisions. Project context solves this by providing one document all agents read first.

Actions

generate

Scan the codebase and create docs/project-context.md from scratch.

Scan targets:

• package.json / Cargo.toml / go.mod / pyproject.toml → tech stack
• tsconfig.json / eslint. / prettier. → code conventions
• /test/ / **/.test. / **/.spec. / **/.cy. → testing patterns
• Existing docs/ → architecture decisions, prior context
• .env.example → environment variables (names only, not values)

Output: docs/project-context.md using templates/project-context-template.md

update

Re-scan codebase and update existing docs/project-context.md. Preserves manual edits in sections marked <!-- manual -->.

init

Write a TODO-filled skeleton to docs/project-context.md for greenfield projects where generate would produce empty sections (no codebase to scan yet).

Behaviour:

• If docs/project-context.md already exists → refuse with error. Use update instead.
• If absent → write templates/skeleton.md to docs/project-context.md.
• Creates docs/ directory if it does not exist.

When to use init vs generate: | Situation | Action | |-----------|--------| | Empty or near-empty project, < 5 source files | init — fill manually | | Established project with real code | generate — auto-derive | | Existing docs/project-context.md is stale | update — re-scan, preserve manual edits |

When to Run

• First time: After mk:bootstrap or mk:docs-init
• After architecture changes: New framework, new database, major refactor
• Periodically: When agents produce inconsistent code (symptom of stale context)

Focus on the Unobvious

The template emphasizes facts agents cannot infer from code:

• WHY a framework was chosen (not just what it is)
• Rejected alternatives ("use Zustand, NOT Redux")
• Anti-patterns specific to this project
• Testing boundaries (what to unit test vs e2e)

Template

See templates/project-context-template.md for the full structure.

Gotchas

• Stale context is worse than no context — re-run update after major architecture changes
• The template asks for "rejected alternatives" — this is the highest-value content. If you chose Zustand over Redux, say WHY
• Don't list every file path — agents can Glob for that. Focus on RULES and CONVENTIONS
• <!-- manual --> sections survive update action but NOT generate — backup before regenerating from scratch
• If project-context.md conflicts with CLAUDE.md, CLAUDE.md wins (it's the higher-priority document)
• init refuses to overwrite an existing docs/project-context.md — this is intentional. Delete the file manually if you truly want to reset to a blank skeleton.

Integration

All agents load docs/project-context.md in their Required Context section. If the file doesn't exist, agents behave exactly as before (graceful degradation).

Files in this skill

• SKILL.md — you are here
• templates/project-context-template.md — output template used by generate action
• templates/skeleton.md — TODO-placeholder constitution template used by init action