The Agent Skills Directory

[SAFE]: The skill establishes a protocol for technical rigor, instructing the agent to prioritize codebase integrity over performative agreement. It encourages pushing back on feedback that is technically unsound, lacks context, or violates YAGNI principles.- [PROMPT_INJECTION]: The skill defines a surface for processing untrusted data from external reviewers. It mitigates indirect prompt injection risks by mandating a verification process (Read-Understand-Verify-Evaluate) before taking action. 1. Ingestion point: GitHub pull request comments accessed via gh api. 2. Boundary markers: Procedural instructions requiring restatement of requirements and verification. 3. Capability inventory: Access to the GitHub API for replies and the ability to modify local source code. 4. Sanitization: Mandatory comparison of external suggestions against the current codebase and project architecture before execution.- [COMMAND_EXECUTION]: Instructions provide a template for using the GitHub CLI (gh api) to reply to specific comment threads, which is a legitimate and expected operation for an agent assisting with PR management.- [NO_CODE]: The skill consists exclusively of instructional documentation in markdown and does not bundle any scripts or executable binaries.

address-pr-comments