The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses various scripts (e.g., branch-audit.sh, safe-push.sh, identity-audit.sh) to execute Git and GitHub CLI (gh) commands. These commands are essential for auditing repository status, managing pull requests, and verifying user identities.
[DYNAMIC_EXECUTION]: The script scripts/main-git-gate.py uses os.execv to perform process replacement, allowing it to act as a filtering proxy for Git commands. This is used to implement granular control over which Git operations are permitted on the main branch, effectively creating a command whitelist.
[PRIVILEGE_ESCALATION]: The scripts/main-guard.py script employs macOS-specific os.chflags to apply the uchg (immutable) attribute to critical Git metadata files (such as HEAD and index). This serves as a platform-level safeguard against accidental branch changes or unauthorized index modifications.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by processing untrusted data from Git commit history and diffs.
Ingestion points: Commit messages, file diffs, and PR metadata processed by scripts/privacy_check.py and scripts/safe-pr.py.
Boundary markers: The skill uses modular Standard Operating Procedures (e.g., references/pr-workflow.md) to scope agent behavior to specific Git tasks.
Capability inventory: Git and GitHub CLI command execution, file system modification, and process replacement via os.execv.
Sanitization: scripts/privacy_check.py implements regex-based scanning for PII and credentials, using a local review mechanism to permit safe exceptions and prevent data exfiltration.

git-workflow