The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from repository files and pull request comments, creating a surface for potential indirect prompt injection attacks.\n
Ingestion points: The skill reads files in the repository to re-baseline (SKILL.md Step 1) and fetches pull request comments using gh pr view --comments during the review loop (SKILL.md Step 8).\n
Boundary markers: The instructions do not define explicit delimiters or warnings to prevent the agent from following instructions embedded within the files or comments it reads.\n
Capability inventory: The skill has access to the Bash tool, allowing it to execute git commands (commit, push), gh commands (create, merge PRs), and arbitrary shell commands defined in the repository's verification gates (SKILL.md Step 6).\n
Sanitization: No explicit sanitization, filtering, or validation of the external content is performed before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill executes shell commands defined in the repository's configuration file, which is the intended behavior for running build and test gates.\n
Evidence: Step 6 in SKILL.md instructs the agent to run install, type-check, lint, test, and build commands defined in the .claude/git-workflow.md file.\n
Evidence: The verification logic in references/merge.md executes the typecheck gate command in a bash loop to verify individual commits.

pr