The Agent Skills Directory

[COMMAND_EXECUTION]: The inspect_pr_checks.py script facilitates interaction with GitHub and Git via subprocess.run calls. These operations are executed using list-based arguments (e.g., ['gh', 'pr', 'view', ...]) rather than raw shell strings, which mitigates the risk of shell command injection.
[PROMPT_INJECTION]: As the skill processes untrusted data from GitHub PR comments and CI logs, it presents a potential surface for indirect prompt injection.
Ingestion points: The inspect_pr_checks.py script fetches review thread contents, PR comments, and GitHub Action logs using the gh CLI tool.
Boundary markers: The SKILL.md instructions enforce a rigid 'Diagnosis Report' structure and 'Progress' template to guide the agent's output and minimize unintended instruction following.
Capability inventory: The skill has the capability to modify repository content, push commits, resolve PR review threads, and post PR comments via the gh CLI and the bundled script.
Sanitization: PR numbers are validated using regular expressions, and GraphQL mutations are performed using variable binding through the gh api -f flag to prevent query injection.
[SAFE]: The skill requires the user to provide an environment where the GitHub CLI (gh) is authenticated with workflow and repo scopes. While these permissions are extensive, they are necessary for the skill's primary function of fixing CI failures and managing PR threads.

gh-fix-ci