The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and performs semantic analysis on untrusted markdown data to find contradictions and stale claims.
Ingestion points: Markdown files located in the wiki/ directory and its subfolders (sources, entities, concepts, synthesis).
Boundary markers: No delimiters or specific instructions are provided to the agent to distinguish between its own system instructions and the content being audited.
Capability inventory: The skill has access to Bash, Read, Write, and Edit tools. It uses these to search the wiki and propose or apply fixes to the files based on its analysis.
Sanitization: The instructions do not specify any validation or sanitization of wiki content, page titles, or wikilinks before they are processed or used in shell commands like grep or the optional qmd tool.

second-brain-lint