The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a variety of complex shell pipelines and local Python scripts (e.g., wiki-sync.py, skill-health.py, tidy.py) to audit the repository. This includes accessing files outside the immediate project root, such as scripts in the ~/.claude/harness/ directory.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a vulnerability surface by ingesting and processing untrusted data from wiki pages, source code comments, and project logs to perform automated actions.
Ingestion points: Processes all markdown files in llmwiki/wiki/, session logs in harness/metrics/, and codebase markers in .py, .js, .ts, and .sh files.
Boundary markers: The skill lacks explicit boundary markers or instructions to ignore embedded commands within the files it parses.
Capability inventory: The agent has the capability to write to index.md, log.md, and the sources/draft/ directory, as well as perform in-place modifications to fix links and resolve index gaps.
Sanitization: There is no explicit sanitization or escaping of content retrieved from the processed files before it is used in reports or follow-up commands.

lint