The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process documents from the raw/ directory. These documents are treated as sources of knowledge, but there are no instructions to sanitize the content or ignore potential instructions embedded within them, which could lead to agent hijacking.
Ingestion points: The skill reads source files from the raw/ directory during the 'Ingest' workflow described in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or guidance to treat external content as data rather than instructions.
Capability inventory: The agent can write to the wiki/ directory, modify project configuration files (e.g., .gitignore, justfile), and execute Python scripts and the zensical tool.
Sanitization: No sanitization or filtering of external content is specified before it is incorporated into the knowledge base.
[COMMAND_EXECUTION]: The skill executes locally provided Python scripts (preprocess_wiki.py and wiki_linter.py) to perform maintenance tasks. These scripts perform file system operations, including directory deletion (shutil.rmtree) and file modifications, based on command-line arguments.
[EXTERNAL_DOWNLOADS]: The templates/justfile configuration uses uvx zensical, which downloads and executes the zensical package from the public PyPI registry at runtime.

llm-wiki