The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to autonomously ingest and act upon pull request review comments and CI status reports from GitHub. This creates a surface for indirect prompt injection where a malicious reviewer could provide crafted feedback to steer the agent's code generation or execution.
Ingestion points: External review comments are fetched and parsed via the triage.dart script (referenced in SKILL.md Step 3 and 4) and GitHub CLI commands.
Boundary markers: The skill includes a 'Critical Mindset' and 'Empirical Skepticism' protocol to verify feedback, but it lacks explicit data delimiters or 'ignore embedded instructions' markers for the ingested comment text.
Capability inventory: The skill possesses significant capabilities including writing to the filesystem, executing git commit, git push, and posting comments via gh pr comment.
Sanitization: No specific sanitization or filtering of the external comment content is documented before it is passed to the agent for remediation.
[COMMAND_EXECUTION]: The skill utilizes shell commands through the GitHub CLI and Dart runtime to automate the pull request lifecycle and manage background tasks.
Evidence: SKILL.md defines workflows using gh pr create, gh run watch, git push, and shell commands like kill <PID> for background process cleanup.

pr-loop