The Agent Skills Directory

[SAFE]: The skill implements several security-positive patterns and constraints:
SHA Verification: Step 10 mandates that CI checks must correspond to the latest pushed commit hash (HEAD_SHA), preventing the agent from relying on stale check results.
Privilege Restriction: Step 13 explicitly forbids using administrative overrides (--admin) or bypassing branch protections during the merge process.
Safety Gates: The skill requires fresh evidence of passing tests, successful builds, and bot approvals before proceeding to merge.
Context Validation: Step 3 prevents the skill from executing on the default/base branch to avoid accidental direct pushes to protected branches.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute standard development utilities such as git, gh (GitHub CLI), jq, and make. These commands are used for PR management, state persistence, and code generation within the user's project environment.
[DATA_EXFILTRATION]: The skill interacts with GitHub's REST and GraphQL APIs to monitor and update pull requests. All network operations are directed at the official GitHub API and are consistent with the skill's primary purpose.
[PROMPT_INJECTION]: The instructions use strong directional language (e.g., "CRITICAL", "MANDATORY") to enforce safety protocols and procedural correctness, rather than attempting to bypass the agent's core safety filters or extract system prompts.

ship