Skip to main content

Scaffold agent context files into a project.

Project description

agentinit

CI PyPI Python Versions

agentinit preview

Scaffold hardened, router-first context files so your AI coding agents stop guessing and start delivering.

Pure Python standard library. No runtime dependencies. Generates 2026-ready manifests for Claude Code, Cursor, Copilot, and Gemini CLI. llms.txt also provides discovery-friendly routing for other tools (for example, Windsurf).

๐Ÿš€ Start in 60 seconds

# 1. Install via pipx (recommended)
pipx install agentinit

# 2. Initialize hardened context in your project
cd your-project
agentinit init --minimal

What it does

Instead of giant, token-heavy instruction files, agentinit implements a hierarchical context strategy. It creates a machine-readable map of your project and enforces autonomy via Hardened Mandates.

# Full scaffold (non-minimal)
your-project/
โ”œโ”€โ”€ llms.txt               # The "robots.txt" for AI (Discovery Index)
โ”œโ”€โ”€ AGENTS.md              # The central hub with Hardened Mandates
โ”œโ”€โ”€ CLAUDE.md              # Claude Code router
โ”œโ”€โ”€ GEMINI.md              # Gemini CLI router
โ””โ”€โ”€ docs/
    โ”œโ”€โ”€ PROJECT.md         # What this project is
    โ”œโ”€โ”€ CONVENTIONS.md     # How to work in it
    โ””โ”€โ”€ STATE.md           # Persistent working memory (AI-readable)

With agentinit init --minimal, only these files are created:

your-project/
โ”œโ”€โ”€ llms.txt
โ”œโ”€โ”€ AGENTS.md
โ”œโ”€โ”€ CLAUDE.md
โ””โ”€โ”€ docs/
    โ”œโ”€โ”€ PROJECT.md
    โ””โ”€โ”€ CONVENTIONS.md

Minimal mode intentionally omits docs/STATE.md, docs/TODO.md, and docs/DECISIONS.md to keep context lean. Use full agentinit init when you want persistent session memory files.

Next steps: Open docs/PROJECT.md and docs/CONVENTIONS.md and fill them in.

Next session (full scaffold): Your agent will find llms.txt, read your rules in AGENTS.md, and follow the mandates to autonomously maintain docs/STATE.md and docs/TODO.md.

Troubleshooting: files not visible to your agent

Some agents only scan tracked files. If your agent says it can't find your context:

  • Track everything: Add the manifests so your agent can see them.

    git add llms.txt AGENTS.md CLAUDE.md GEMINI.md docs/
    git add .agents/  # if you added extras
    
  • Verify ignores: Run git status --ignored to see if your .gitignore is hiding them.

Minimal .gitignore exceptions
!llms.txt
!AGENTS.md
!CLAUDE.md
!GEMINI.md
!docs/PROJECT.md
!docs/CONVENTIONS.md
!docs/TODO.md
!docs/DECISIONS.md
!docs/STATE.md
!.agents/
!.agents/**

๐Ÿ› ๏ธ Add extras (Skills, MCP, Personality)

Agents can do more than just read conventions. Inject modular, hardened resources directly into your project.

# View available resources by type
agentinit add skill --list
agentinit add mcp --list

# Add specific capabilities
agentinit add skill code-reviewer
agentinit add security
agentinit add soul "YourAgentName"

Resources in .agents/ are automatically linked in AGENTS.md and use Imperative Mandates (MUST ALWAYS, MUST NEVER) to ensure compliance and zero-sycophancy.


๐Ÿšฆ Keep your context clean (Status & CI)

Token limits matter. agentinit includes a validator to prevent context bloat, broken links, and missing information.

# View line budgets, broken references, and missing details
agentinit status

# Perfect for CI (exits non-zero on violations)
agentinit status --check

๐Ÿค– AI Prompt: Fill the docs fast

After running agentinit init, paste this to your favorite AI agent:

Read the entire repository. Fill in docs/PROJECT.md and docs/CONVENTIONS.md using only facts found in the repo (package files, configs, source code, CI). Do not invent commands. If information is missing, write TODO: <what's needed>. Update docs/STATE.md and docs/TODO.md to reflect our current progress. Do not modify any other files.


๐Ÿ“š Advanced Usage & Commands

Core Commands

  • agentinit init โ€” Add missing files to an existing directory.
  • agentinit minimal โ€” Shortcut for init --minimal.
  • agentinit status โ€” Check health, line budgets, and broken links.
  • agentinit add <type> <name> โ€” Add modular resources (skills, mcp, security, soul).
  • agentinit new <project> โ€” Create a new directory and scaffold files.
  • agentinit remove โ€” Safely remove or archive agent files.

Common Flags

  • --detect โ€” Auto-detect stack and commands from package files (e.g., package.json, pyproject.toml).
  • --yes / -y โ€” Runs non-interactively and skips the wizard.
  • --purpose "<text>" โ€” Prefill the project purpose non-interactively.
๐Ÿ’ก Why Hardened Context?
  • Agent Autonomy: Explicit mandates (YOU MUST ALWAYS read state.md) transform the agent from a chatbot into a disciplined project maintainer.
  • AI-Discovery Index: llms.txt ensures any tool (Cursor, Windsurf, Claude) immediately understands your project map.
  • Progressive Disclosure: Keeps the context window high-signal by loading deep details only when relevant.
  • Zero Sycophancy: Mandates force the agent to skip "I'd be happy to help!" and jump straight to the technical solution.
โš™๏ธ Installation & Development

Requires Python 3.10+.

# Install (pipx is recommended)
pipx install agentinit

# Update
pipx upgrade agentinit

Development

pip install -e . --group dev
python3 -m pytest tests/ -v

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

agentinit-0.3.3.tar.gz (46.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

agentinit-0.3.3-py3-none-any.whl (44.8 kB view details)

Uploaded Python 3

File details

Details for the file agentinit-0.3.3.tar.gz.

File metadata

  • Download URL: agentinit-0.3.3.tar.gz
  • Upload date:
  • Size: 46.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for agentinit-0.3.3.tar.gz
Algorithm Hash digest
SHA256 24b33aa90ec50272515c1c2a3c4e85ece18bea939828ce08246be6c547501295
MD5 7f580f644fda0122f9745a07b3c66006
BLAKE2b-256 66701bd8999f6438589efcce0e352da522da57918e6fb665edfc206b4b4a2b5b

See more details on using hashes here.

Provenance

The following attestation bundles were made for agentinit-0.3.3.tar.gz:

Publisher: publish.yml on Lucenx9/agentinit

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file agentinit-0.3.3-py3-none-any.whl.

File metadata

  • Download URL: agentinit-0.3.3-py3-none-any.whl
  • Upload date:
  • Size: 44.8 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for agentinit-0.3.3-py3-none-any.whl
Algorithm Hash digest
SHA256 357578dbb6201b444e08e30f20ab530b8aba78fbfe4db4b26f9550377ffcd9b3
MD5 aade35b4b0786021fd2c6d5c9c05c6cc
BLAKE2b-256 5a45f4895e2fac202f506f2db1523e4499a04e9e3bd5f1b30ff6f3c643851d94

See more details on using hashes here.

Provenance

The following attestation bundles were made for agentinit-0.3.3-py3-none-any.whl:

Publisher: publish.yml on Lucenx9/agentinit

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page