The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by reading repository files such as CONTRIBUTING.md and .github/PULL_REQUEST_TEMPLATE.md to identify pre-PR steps. An attacker with write access to the repository could potentially insert malicious instructions or commands into these files to be processed by the agent.\n
Ingestion points: Repository documentation files including CONTRIBUTING.md and PULL_REQUEST_TEMPLATE.md.\n
Boundary markers: None; the agent processes the file content directly to extract actionable steps.\n
Capability inventory: Shell command execution via gh CLI, local script execution (.agents/skills/pre-push/run-checks.sh), and generic project tooling (npm, pytest, etc.).\n
Sanitization: The skill mitigates risks by listing the discovered steps and requiring the user to approve which ones to run before execution.\n- [COMMAND_EXECUTION]: The skill executes shell commands to perform its primary function, including gh pr create and a vendor-specific pre-push script located at .agents/skills/pre-push/run-checks.sh. It also facilitates the execution of build and test commands (e.g., npm test, pytest) found in the repository's own documentation.

pr-create