The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill reads data from the local project (including source code and existing markdown files) and incorporates it into agent-accessible documentation like reports, indexes, and glossaries.
Ingestion points: The scan and discover functions in lib.ts read project code files to identify technical patterns and definitions. The generateIndex and generateGlossary functions read existing markdown documentation within the docs/knowledge directory.
Boundary markers: Ingested content is placed within Markdown templates using headers and blockquotes, though it lacks explicit instructions to the agent to ignore embedded commands within that content.
Capability inventory: The skill performs filesystem writes (mkdir, writeFile) to manage the knowledge base structure and documents. It does not perform network operations or active code execution.
Sanitization: The create function in lib.ts sanitizes document names and category paths using regex (/[^a-zA-Z0-9\-_]/g and /[^a-zA-Z0-9\-_/]/g) to prevent path traversal attacks.

knowledge-base