The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to utilize the exec tool for running various CLI utilities, including gbrain, bun, jq, ls, and test runners to perform audits and verify skill integrity.
[DYNAMIC_EXECUTION]: The workflow requires the agent to dynamically generate TypeScript/JavaScript code in scripts/ directories and execute these scripts during the testing and verification phases (e.g., Phase 4 and Phase 7).
[INDIRECT_PROMPT_INJECTION]: As a tool designed to ingest, audit, and improve other skill definitions (SKILL.md) and their associated code, the skill possesses an attack surface for indirect prompt injection where instructions embedded in the analyzed data could attempt to influence the agent's behavior.
Ingestion points: Reads and parses SKILL.md, associated scripts, and routing-eval.jsonl files (SKILL.md).
Boundary markers: The instructions do not explicitly mandate delimiters or sanitization for the content being processed during the audit.
Capability inventory: Uses exec, write, edit, and read tools to modify and run code (SKILL.md).
Sanitization: There are no specific instructions provided for escaping or validating the content of the ingested skill files before processing or evaluation.

skillify