The Agent Skills Directory

Indirect Prompt Injection Surface: The skill ingests untrusted issue bodies and comments from external trackers (Step 2). It mitigates this risk through a strict policy of treating external content as data rather than instructions and requiring human confirmation for all public posts (Golden Rule 2). \n- Command Execution: To perform its functions, the skill utilizes CLI tools such as gh for tracker interaction and git for repository analysis (Steps 2 and 6). It includes specific guidelines to prevent shell injection by using temporary files for message bodies instead of direct command-line arguments (Step 6). \n- Security Signal Screening: The skill implements a critical screening phase (Golden Rule 8) to detect remote code execution mentions or credential exposure in issue reports, halting the workflow to prevent accidental public disclosure of vulnerabilities. \n- Maintainer Oversight: All proposed actions are presented as drafts for user review and confirmation before execution, ensuring the AI agent remains under human control for all state-changing tracker operations.

triage