The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local helper script (scripts/guard.mjs) to perform security-critical path validation. This script ensures that all file operations remain within the user-defined target directory and generates SHA-256 hashes to verify the integrity of files before they are modified.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted 'material' (notes, links, or data dumps) and reads from an existing knowledge base, which creates a vulnerability surface for indirect prompt injection.
Ingestion points: Untrusted material provided as input and existing documents within the target knowledge base.
Boundary markers: The skill instructions contain explicit 'Rules' that mandate ignoring any instructions found within the data, treating all input strictly as content to be landed rather than commands to be followed.
Capability inventory: The agent is empowered to write to the local file system and execute the included Node.js validation script.
Sanitization: All destination paths are sanitized via the guard.mjs script, which blocks absolute paths, directory traversal ('..'), and symlink escapes to ensure edits are confined to the intended root.

re0-import