Init Project

Automatically configure Claude Code best practices for new projects: directory skeleton, general-purpose agents, hooks, and AGENTS.md (with CLAUDE.md as a thin @AGENTS.md stub).

Execution Flow

1. Phase 1: Generate Skeleton (deterministic script) → Run scripts/init-skeleton.sh to create directory structure + boilerplate files → Output which files were created

2. Phase 2: Interactive AGENTS.md Fill (LLM + AskUserQuestion) → Process <!-- init-project: 待填充 --> placeholders section by section → For each section: read codebase → generate draft → AskUserQuestion to confirm → write → User can reply "skip" to skip any section → CLAUDE.md is just a stub (@AGENTS.md + write-routing rule); never edited in Phase 2 → Detailed guide templates in details/agents-md-sections.md

3. Phase 3: Profile Overlay (optional) → AskUserQuestion: "Want to overlay an additional profile? Currently supported: research" → If research selected → run scripts/init-research-profile.sh → Profile details in details/research-profile.md

4. Output Summary → List all generated/modified files → Suggest next steps: review content, git add, start developing

Phase 2 Section Processing Strategy

Section	Auto-exploration	User prompt
Project overview	Read README, pyproject.toml, package.json	"One-line description of this project's core goal?"
Models / Key dependencies	Detect imports, API keys, model configs	"What models or external services does this project depend on?"
Directory structure	ls + read key file docstrings	"Does this directory layout look correct? Anything to adjust?"
Dev workflow	Detect CI, Makefile, scripts/	"Use the default brainstorming→plans→dev→verify flow?"
Dev guide	Detect venv, .env, Dockerfile	"Any special environment setup steps?"
Context Loading (routing table)	List .claude/rules/*.md + scan opening summaries	"Confirm (task domain → rule) mapping — entry point for Codex / external agents"
Always Do (project-specific)	Read rules/, lint config	"Any cross-module consistency requirements?"
Ask First	Scan core files (interfaces, config)	"Which files/dirs require confirmation before modifying?"
Never Do	Detect third_party/, .env	"Any absolute don't-touch conventions?"
Progressive disclosure setup	Scan docs/knowledge/ + .claude/rules/	Ensure rules reference knowledge files inline; ask about missing refs

Constraints

• Idempotent: existing files are never overwritten, only gaps are filled
• No auto git add/commit: user decides when to commit
• No existing content modification: only <!-- init-project: 待填充 --> placeholders are touched
• Scripts run standalone: init-skeleton.sh and init-research-profile.sh work independently of the skill

References

• Skeleton file manifest: details/skeleton-manifest.md
• AGENTS.md section guide: details/agents-md-sections.md
• Agent templates: details/agent-templates.md
• Research profile: details/research-profile.md