The Agent Skills Directory

[COMMAND_EXECUTION]: Orchestrates project management and CI/CD tasks using standard CLI tools including npm, git, and the GitHub CLI (gh). These operations are confined to the repository scope and are necessary for the skill's stated engineering management purpose.- [REMOTE_CODE_EXECUTION]: Utilizes the platform's native Agent tool to spawn subagents for task parallelization and delegation. This follows standard platform patterns for autonomous operations and does not involve execution of untrusted remote scripts.- [DATA_EXFILTRATION]: Communicates with GitHub APIs via the gh tool to manage issue and pull request metadata. All network activity is directed toward official GitHub endpoints; no exfiltration of sensitive host data or credentials was detected.- [PROMPT_INJECTION]: Employs strict internal 'Blocking Gates' and enforcement tags (e.g., <tdd_enforcement>, <em_rules>) to maintain workflow integrity. These instructions guide the agent's behavior to follow rigorous development standards rather than attempting to bypass safety filters.- [SAFE]: Incorporates sophisticated defensive measures, such as the 'Validation Engineer' and 'Claim Verification' gates, which require hard evidence (official documentation, source code, or reproduction) before the agent acts on data from GitHub issues. This proactive design effectively mitigates risks associated with indirect prompt injection and platform hallucinations.

context-mode-ops