The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves text chunks from an external index containing USPTO MPEP, 35 USC, and 37 CFR content and injects them into the agent's context. While the source is highly reputable, the ingestion of any external data creates a surface where embedded instructions could potentially influence the agent's behavior.
Ingestion points: The search_mpep and get_mpep_section functions in mpep_search.py retrieve text from a local FAISS/BM25 index.
Boundary markers: No explicit delimiters or instructions are applied to the retrieved text within the script.
Capability inventory: The skill is limited to read operations and has no capabilities for file-write, network operations, or subprocess execution.
Sanitization: The script does not perform sanitization or filtering of the retrieved legal text before returning it to the agent.
[EXTERNAL_DOWNLOADS]: The skill implementation in mpep_search.py references a setup script (install.py) that is responsible for downloading and indexing the patent corpus. While this script is not executed automatically by the skill, it indicates a dependency on external content fetched during installation.
[DYNAMIC_EXECUTION]: The script mpep_search.py dynamically modifies the Python path (sys.path) at runtime to import modules from a calculated directory structure (Path(__file__).parent.parent.parent.parent). This is a common pattern for local development but represents a dynamic loading mechanism.

mpep-search