Metadata-Version: 2.4
Name: tulki
Version: 0.3.0
Summary: Local AI-agent readiness doctor, merged-PR audit, policy generator, and coding-agent hook.
Author: Sanjok Bhatta
License-Expression: LicenseRef-Proprietary
Keywords: ai-agents,mcp,readiness,repository-scanner,claude-code,hooks
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: PyYAML>=6.0
Dynamic: license-file

# Tul-KI: Doctor, Audit, Policy Init, Local Agent Hook

Local tooling for teams adopting AI coding agents. Nothing here requires
the Tul-KI GitHub App, hosted runtime, or receipt signing keys. `doctor`,
`policy init`, and `hook` run fully offline; `audit` talks only to the
GitHub API with a read-only token you supply — no data leaves your machine
in any command.

Generate a starter policy from your repository structure, then assess:

```bash
tulki policy init --repo .
tulki doctor --repo .
tulki doctor --scan-mcp
```

Audit your last 90 days of merged PRs — how many were agent-authored, how
many touched sensitive paths, and how many merged on an approval given
before the final push (so the approver never saw the shipped diff):

```bash
tulki audit --repo OWNER/REPO --days 90            # needs GITHUB_TOKEN
tulki audit --repo OWNER/REPO --format markdown --out audit.md
```

Put a guardrail on your own coding agent: the Claude Code PreToolUse hook
evaluates `tulki-policy.yml` before the agent edits a file. Allowed paths
stay silent, approval-required paths ask the human with the policy rule
and approver groups, and blocked paths are denied.

```bash
tulki hook claude-code --print-config
```

Merge the printed snippet into `.claude/settings.json` and the hook is
active.

This package is the public local wedge only. The private Tul-KI
enforcement product, GitHub App, receipt signing flow, deploy gate, hosted
runtime, and operator runbooks are not included in this distribution.
