Safety
How madtea keeps the token out of argv, history, and errors, and how its write-side guards fail closed.
| Doc | What it covers |
|---|---|
| credential-safety.md | The threat model and the storage backends that keep the token out of argv, history, and errors. |
| safe-body-edits.md | Anchored, optimistic-locked edits that can’t trample an issue or PR body. |
| safety-hooks.md | The seven Claude-Code-only bash hooks that steer agents away from credential-leaking raw git. |
| wrong-repo-protection.md | The fail-closed guard that refuses an MCP write to the wrong repo in a multi-repo workspace, and how MADTEA_CALLER_REPO / owner_repo make it reliable. |
Part of the madtea documentation tree.
- Credential Safety How madtea stores your Forgejo/Gitea token locally and what the “credential-safe git wrappers” claim actually guarantees.
- Safe body edits madtea has no command that overwrites an issue body or pull-request body wholesale.
- Safety hooks The Claude Code plugin installs seven PreToolUse hooks that intercept tool calls before they run — they don’t change git itself, only what the agent is permitted to run.
- Wrong-repo fail-closed guard A repo-scoped MCP tool called without an explicit owner_repo resolves its target repo from the MCP server process’s working directory — not from any cd the agent did.