The Agent Skills Directory

[DATA_EXFILTRATION]: The skill reads multiple markdown files from the docs/kb/ directory and CLAUDE.md to synthesize answers. While these are project documentation files, this constitutes a data access pattern that could be leveraged if combined with network operations, though no network activity is present in this skill.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the knowledge base files it processes.
Ingestion points: Reads content from any .md file within docs/kb/ and the CLAUDE.md catalog.
Boundary markers: Absent. There are no explicit instructions to the agent to ignore or delimit potentially malicious instructions contained within the KB files being synthesized.
Capability inventory: The skill has the ability to read files, create new markdown files in docs/kb/, and modify existing project files such as CLAUDE.md, _index.md, and _log.md.
Sanitization: Absent. The skill does not validate or sanitize the content retrieved from the KB files before using it in synthesis or writing it back to new files.

kb-query