The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates multiple command-line utilities, including git for repository management, gh for GitHub interactions, and herdr for session pane control.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect injection as it processes untrusted data from pull request comments and the session history. * Ingestion points: Pull request comments via gh pr view and the entire session history. * Boundary markers: No specific delimiters are used to wrap external content. * Capability inventory: Execution of shell commands (git, gh, herdr) and file writing to the knowledge base. * Sanitization: Includes explicit instructions to redact credentials and personal data from output. * Mitigation: The risk is significantly reduced by the mandatory use of structured user questions for all judgmental decisions, ensuring a human-in-the-loop.
[DATA_EXFILTRATION]: The skill reads repository state and identifies local files like .env. This is performed specifically to warn the user about potential data loss during worktree deletion. The skill includes explicit instructions to redact sensitive information before any external data transfer to the specified knowledge base.
[SAFE]: The skill exhibits safe operational patterns, including the use of force-with-lease for git pushes and the redaction of personal data in lessons.

settle