The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting and processing untrusted text from wiki pages.
Ingestion points: Wiki page markdown files are read by scripts/extract_page_claims.py to identify claim candidates.
Boundary markers: No explicit delimiters or 'ignore embedded instructions' markers are used when passing extracted data to the cogni-claims:claims skill.
Capability inventory: The skill uses Bash to execute local scripts and the Skill tool to invoke external claim verification; it also performs file writes to the wiki root.
Sanitization: Content is passed through a strip_markdown function which removes formatting but does not filter for malicious instruction content.
[COMMAND_EXECUTION]: The skill uses the Bash tool to run local Python scripts (extract_page_claims.py and resweep_planner.py) for data extraction and report generation. These commands are executed within the context of the local file system and are used to manage the audit workflow.

wiki-claims-resweep