| name | analyze-project |
| description | Generate or refresh a multi-agent project context document and execution-plan authoring prompt from repository analysis. Use when bootstrapping multi-agent orchestration, updating architecture/ownership/contracts guidance, or preparing a reusable planning prompt. Support explicit output path (`OUTPUT=<path>` or plain-language path request); otherwise detect a sensible default path from the project structure. |
| allowed-tools | ["Read","Write","Edit","Grep","Glob","Bash"] |
| argument-hint | [TYPE=project-context|authoring-prompt] [OUTPUT=<path>] |
Multi-Agent Project Context
Create a high-quality multi-agent planning artifact from real repository structure and conventions.
Inputs
TYPE (optional):
project-context (default): build/update the project context foundation document.
authoring-prompt: build/update the execution-plan authoring prompt.
OUTPUT (optional): explicit file path for the generated document.
- Honor an explicit user path even if it differs from defaults.
Output Path Resolution
- Use
OUTPUT when supplied.
- Otherwise resolve a default path with:
bun .claude/skills/analyze-project/scripts/resolve-output-path.ts --kind <project-context|authoring-prompt>
- Create missing parent directories before writing.
- Overwrite the target file unless the user explicitly asks for append mode.
Workflow
- Determine artifact type (
TYPE) and output path.
- Load repository context before drafting:
AGENTS.md
CLAUDE.md
README.md
documentation/GETTING_STARTED.md (if present)
documentation/ARCHITECTURE.md (if present)
documentation/TESTING.md (if present)
.claude/agents/*.md
.claude/rules/*.md
.claude/skills/**/SKILL.md (scan names/descriptions; read deeply only when relevant)
- Map facts into the required template:
- For
project-context, follow references/project-context-template.md.
- For
authoring-prompt, follow references/authoring-prompt-template.md.
- Replace placeholders with concrete repository values.
- Keep scope realistic:
- Prefer existing role names from
.claude/agents/.
- Use real file paths.
- Use actual verification commands (Bun scripts in this repository).
- Write the final document to the resolved path.
Quality Bar
- Keep instructions executable and specific.
- Keep section structure stable so downstream agents can parse it.
- Avoid vague language when acceptance criteria or risk mitigation are required.
- State assumptions explicitly where repository evidence is missing.
- Do not invent unavailable tools, services, or environments.
Validation Checklist
Run these checks before finishing:
- Output file exists and is non-empty.
- Every path in the document exists now or is clearly marked as proposed new work.
- For
project-context: sections 1-9 are present.
- For
authoring-prompt: sections A-G plus self-review are present.
- Commands and package-manager usage align with repository standards (
bun).
References
- Project context template:
references/project-context-template.md
- Authoring prompt template:
references/authoring-prompt-template.md