agentleFS
Sign inStart free
claude-worldclaude-world/director-mode-lite/skills/project-init/SKILL.md

project-init

Skill80 starsChanged 8 months agoMITView on GitHub

Set up concise, guidance-first Claude/Codex/Grok project context after inspecting the repository. Use on first setup, after cloning, or when the user runs /project-init.

We split this skill into the rules it gives an agent

7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 17 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create the smallest useful project context without adding permission gates, blocking hooks, or mandatory process rules.

    line 9
  2. Distinctive:

    Read existing CLAUDE.md, AGENTS.md, README, manifests, scripts, and CI.

    line 14
  3. Distinctive:

    If the stack or desired workflow cannot be inferred safely, ask one focused question instead of inventing it.

    line 18
  4. Distinctive:

    Keep permanent guidance focused on facts useful in most sessions:

    line 23
  5. Very specific:

    Use the Director managed block in both CLAUDE.md and AGENTS.md so all three CLIs can find .director-mode/GUIDANCE.md.

    line 45
  6. Distinctive:

    Check the selected CLI surfaces:

    line 51
  7. Very specific:

    Run scripts/verify-install.sh when available.

    line 60

“Common” is measured: kept by at least 6 in 10 of the 17 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Project initialization

Create the smallest useful project context without adding permission gates, blocking hooks, or mandatory process rules.

1. Inspect before writing

  • Read existing CLAUDE.md, AGENTS.md, README, manifests, scripts, and CI.
  • Detect the language, framework, build, test, lint, and development commands from repository evidence.
  • Preserve existing user guidance and unrelated configuration.
  • If the stack or desired workflow cannot be inferred safely, ask one focused question instead of inventing it.

2. Add concise shared context

Keep permanent guidance focused on facts useful in most sessions:

# Project name

## Purpose
[What this repository delivers]

## Stack and structure
- [Language/framework]
- [Important entry points]

## Commands
- dev: [verified command]
- test: [verified command]
- build: [verified command]
- lint: [verified command]

## Local conventions
- [Only repository-specific conventions supported by evidence]

Use the Director managed block in both CLAUDE.md and AGENTS.md so all three CLIs can find .director-mode/GUIDANCE.md. Put detailed, conditional procedures in skills rather than expanding the permanent files.

3. Verify native assets

Check the selected CLI surfaces:

Claude: .claude/skills/ and .claude/agents/
Codex:  .agents/skills/ and .codex/agents/
Grok:   shared Claude-compatible skills and .grok/agents/
Relay:  .director-mode/bin/director-relay

Run scripts/verify-install.sh when available. Do not require every optional tool merely to declare setup complete.

4. Optional integrations

MCP servers, hooks, TDD, autonomous loops, and extra agents are choices, not defaults. Suggest one only when it materially helps the current repository and explain its effect. Install or register it only after the user requests it.

In particular, never register a Stop hook, deny rule, validator, security guard, or background tracker as part of ordinary project initialization.

5. Report the result

Summarize detected facts, files created or updated, validation actually run, and optional next steps. A useful default next step is director-mode or workflow; use session-relay when another CLI should take over.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free