The Agent Skills Directory

[COMMAND_EXECUTION]: Orchestrates local shell commands using git and gh (GitHub CLI) to automate the lifecycle of issue resolution, including branch creation and status tracking. Commands are derived from the repository state and the issue number provided.
[SAFE]: Implements a human-in-the-loop requirement in Phase 3c, forcing the agent to pause and receive explicit user approval of the generated implementation plan before writing any code or making commits. This effectively mitigates the risk of the agent autonomously acting on malicious content found within an issue.
[PROMPT_INJECTION]: Ingests potentially untrusted content from GitHub issues during the analysis and planning phases. The risk is mitigated by the structured orchestration and the final review step.
Ingestion points: External issue data is fetched via the gh tool in Phase 2 and Phase 1 (via sub-skill).
Boundary markers: No specific delimiters are used to wrap the untrusted issue content within the orchestration logic, but the subsequent human review phase serves as a functional control.
Capability inventory: Possesses Bash execution, file read/write, and git repository write/push capabilities.
Sanitization: Sanitizes GitHub issue titles into alphanumeric slugs for branch naming using sed and tr to ensure they do not interfere with shell execution.

resolve-gh-issue