The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the development lifecycle by executing a wide range of system commands, including git, gh, python3, cargo, dotnet, and go to manage branches, run tests, and create pull requests.\n- [REMOTE_CODE_EXECUTION]: The skill interacts with remote repositories on GitHub via git fetch, git pull, and git push operations. These are essential for its purpose but involve interacting with remote code states.\n- [EXTERNAL_DOWNLOADS]: It fetches code updates and project dependencies from GitHub, which is recognized as an established and well-known service.\n- [DATA_EXFILTRATION]: Project source code and metadata are pushed to GitHub repositories. The push command templates in the skill and examples contain a hardcoded GitHub username (KoshelevDV), which could lead to code being pushed to an unintended destination if not correctly configured by the user.\n- [PROMPT_INJECTION]: The skill processes untrusted data from git diff outputs and project documentation (AGENTS.md) to populate prompts for review subagents. Mandatory Evidence Chain: 1. Ingestion points: git diff output and AGENTS.md read via shell commands in SKILL.md. 2. Boundary markers: Data is delimited using headers such as PROJECT_CONTEXT and DIFF. 3. Capability inventory: Extensive command execution, file system access, and network operations (PR creation). 4. Sanitization: Absent; the skill relies on the agent's internal verification of findings.\n- [SAFE]: The use of cron jobs for 'anti-freeze' logic is a functional persistence mechanism explicitly documented as a way to ensure the long-running pipeline resumes correctly after asynchronous tasks.

Full Cycle Developer