The Agent Skills Directory

[COMMAND_EXECUTION]: The lint_wiki.py and rebuild_open_questions.py scripts utilize subprocess.run to execute companion scripts (config_bump.py, wiki_index_update.py, and lint_wiki.py) located within the same plugin directory. These operations use hardcoded relative paths and are used for routine maintenance tasks such as updating configuration values and reflowing index files.
[PROMPT_INJECTION]: The skill performs a semantic audit (Step 4 in SKILL.md) where the LLM reads and reasons about the content of various wiki pages to identify contradictions and type drift. This constitutes an indirect prompt injection surface as malicious instructions embedded within wiki pages could potentially influence the model's reasoning during the audit.
Ingestion points: Reads arbitrary markdown files under the wiki/ directory via iter_pages() in lint_wiki.py.
Boundary markers: None explicitly defined in the provided workflow to separate wiki content from audit instructions.
Capability inventory: The agent has Read, Write, Edit, and Bash capabilities, which are used to generate reports and update wiki metadata.
Sanitization: No evidence of content escaping or validation before passing page bodies to the LLM for semantic analysis.

wiki-lint