The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): The skill contains behavioral instructions to avoid 'performative agreement' and prioritize technical reasoning. These instructions are persona-related and do not attempt to bypass core AI safety guardrails or extract system prompts.
[COMMAND_EXECUTION] (SAFE): The skill suggests using the grep command to verify code usage (YAGNI checks). This is a standard, read-only development operation and is not combined with any malicious or privileged escalation patterns.
[DATA_EXFILTRATION] (SAFE): There are no patterns suggesting the exfiltration of sensitive data. Mention of the GitHub API (gh api) is contextual to the skill's purpose (replying to PR comments) and does not involve unauthorized data access or credential leakage.
[INDIRECT_PROMPT_INJECTION] (SAFE): The skill is designed to handle external feedback, which is an ingestion point for potentially untrusted data. However, the skill explicitly mitigates this risk by instructing the agent to 'be skeptical', 'verify against codebase reality', and 'push back with technical reasoning', effectively providing a safety layer against adversarial feedback.
[OBFUSCATION] (SAFE): No obfuscated content, Base64 encoding, zero-width characters, or homoglyphs were detected in the skill.

receiving-review