The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple local Python scripts (e.g., concept-store.py, backlink_audit.py, wiki_index_update.py) within the plugin's directory to manage wiki state and process distilled information.
[DYNAMIC_EXECUTION]: The skill dynamically resolves its own script paths using a helper script (resolve-wiki-scripts.sh) to support versioned utility directories within the plugin root.
[INDIRECT_PROMPT_INJECTION]: The skill processes claim data derived from external source pages, which serves as a potential attack surface. 1. Ingestion points: Reads claims from wiki/sources/<slug>.md and question nodes. 2. Boundary markers: Uses structured frontmatter blocks and machine-owned sentinels to delimit content. 3. Capability inventory: Uses Bash for script execution and Write for file modifications. 4. Sanitization: Employs a sanitize_summary function to filter content before index integration.

knowledge-distill