The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from a user-specified codebase to identify 'surprising' or 'foundational' facts. This creates a surface for indirect prompt injection where malicious instructions hidden in code comments or documentation could be promoted into the agent's persistent memory (Posts). While the skill requires user approval before posting, the agent itself could be influenced during the discovery phase.
[DATA_EXFILTRATION]: The skill executes git remote get-url origin to identify the repository. While standard for project identification, git remote URLs can occasionally contain hardcoded credentials (e.g., https://token@github.com/...). The skill explicitly instructs the agent to 'Never include secrets,' which serves as a mitigation for this risk.

introduce