The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various Git and GitHub CLI (gh) commands to perform branch operations.
Evidence: The skill instructions include sequences such as git checkout, git pull, git merge, git push, and gh pr create.
Risk Mitigation: The skill explicitly requires a human-in-the-loop for all such operations, mandating confirmation before every individual command. It also contains strict prohibitions against dangerous history-modifying flags like --force or --amend.
[COMMAND_EXECUTION]: The skill interacts with external issue trackers via CLI or API to close tickets.
Evidence: Mentions checking for tracker CLI availability and performing ticket closure based on URLs found in local seed files.
Risk Mitigation: Includes a preflight check and clear error handling if the necessary tools are missing.
[PROMPT_INJECTION]: The skill processes untrusted or external data (ticket URLs, specification files, and Git history) which could technically serve as an indirect injection vector.
Ingestion points: Reads ledger.md, spec.md frontmatter, and External: ticket links from seed files.
Boundary markers: Relies on structured YAML frontmatter and specific file paths.
Capability inventory: Can perform network pushes and ticket closures.
Sanitization: Mitigation is handled via the 'Anti-Sycophancy' logic which instructs the agent to analyze the state of the repo and challenge user decisions if they conflict with evidence (e.g., merging with uncommitted changes).

finish