The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script located at scripts/graph-audit to validate the integrity of the queue.md file. It also utilizes system search utilities such as rg (ripgrep) or grep to perform keyword searches across the file system.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from markdown files within the wiki/ and hub/ directories to synthesize answers for the user.
Ingestion points: The skill reads index.md, various files in wiki/ folders, and hub/ section files as its primary data sources.
Boundary markers: The instructions require the agent to "quote the exact supporting passage" and "cite the exact file," which serves as a logical constraint, though no explicit technical delimiters are defined for the data interpolation.
Capability inventory: The agent can read files, append synthesized notes to queue.md, and execute the scripts/graph-audit validation script.
Sanitization: The skill mandates running scripts/graph-audit on the queue.md file to detect and fix violations before any new entry is considered durable.

recall