The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several TypeScript scripts (e.g., scripts/run-trigger-eval.ts, scripts/audit-skill-links.ts) designed to be run using the bun runtime. The evaluation harness programmatically invokes the claude CLI tool to perform live routing tests. These executions are scoped to local repository maintenance and testing tasks as described in the skill documentation.
[INDIRECT_PROMPT_INJECTION]: The run-trigger-eval.ts script is designed to process external corpus files (evals/routing.json) containing natural language prompts which are then sent to an LLM for evaluation. While this represents a data ingestion surface, the script is a developer testing tool with built-in corpus validation logic, and the interaction is restricted to a specialized routing probe context.
[SAFE]: No malicious patterns were detected. The skill instructions focus on repository governance and instruction quality. There are no attempts to exfiltrate credentials, establish persistence, or bypass agent safety constraints. The local file paths found in log files (evals/runs/*.json) are consistent with vendor developer environments.

agent-instructions