The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the Bash tool to execute various Python scripts and inline Python logic for manifest building, metric recording, and output verification. It interpolates parameters like --project-path and --knowledge-root into these commands, which is a common pattern for orchestration but requires trust in the input sources.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes data from external sources (captured in the wiki/ directory) to generate report drafts.
Ingestion points: Reads from wiki/sources/*.md, wiki/syntheses/*.md, and wiki/index.md (via Step 0 and Step 4).
Boundary markers: The skill instructions do not specify any delimiters or safety instructions (e.g., 'ignore embedded instructions') when the wiki-composer agent processes these files.
Capability inventory: The skill possesses the Bash, Write, Read, and Task capabilities.
Sanitization: While it uses json.dumps for structural metadata serialization, it does not explicitly sanitize the content of the wiki pages before they are used for drafting.

knowledge-compose