The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to locate and run its internal scripts.
Evidence: The documentation instructs the agent to use find ~/.claude/plugins/cache to resolve the path to the search script and execute it via $PATENTS_SCRIPT.
Evidence: The setup process involves running scripts/search setup <api-key>, which passes credentials as a command-line argument.
[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes large volumes of patent text from an external API, which could be used as a vector for indirect instructions.
Ingestion points: Data entering via the results object from api.valyu.ai, specifically the content field which contains full-text patent claims and descriptions (SKILL.md).
Boundary markers: Absent. There are no explicit instructions for the agent to treat the content field as data only or to ignore embedded instructions.
Capability inventory: The skill utilizes subprocess execution via scripts/search and scripts/search.mjs (SKILL.md).
Sanitization: No sanitization or escaping of the remote API content is specified before the agent processes the results.

patents-search