The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains a management script (scripts/secretary.ts) that executes Git commands to generate status briefings and create checkpoint tags. These operations are restricted to project-specific metadata and are intended for workflow orchestration and rollback capability.\n- [DYNAMIC_EXECUTION]: The skill dynamically selects and loads Markdown instruction files ('mode playbooks') based on user intent. This orchestration pattern is designed to limit context window overhead and does not involve the execution of untrusted runtime code.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external inputs to triage and route tasks. It mitigates injection risks by requiring Socratic adversarial reviews and mandatory cryptographic hash approvals for any mutations, ensuring high-stakes actions are reviewed by a human.\n- [DATA_EXFILTRATION]: Access to files in the user's home directory (e.g., ~/.agents/identity/) is used to store and retrieve agent personas and global configuration, which is consistent with the skill's stated purpose as a context and identity manager.\n- [SAFE]: No malicious patterns, remote code execution from untrusted sources, or exfiltration to third-party domains were detected. The skill improves security posture by introducing mandatory human-in-the-loop validation for irreversible actions.

secretary