The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes patent claim text from the Zhihuiya database, which acts as an untrusted data source.
Ingestion points: The claims field in the JSON response from the /zhihuiya/claimDataTranslated API endpoint, processed in scripts/zhihuiya_claim_translated.py.
Boundary markers: No specific delimiters or "ignore embedded instructions" markers are used when presenting the retrieved claim text to the agent or user.
Capability inventory: The skill possesses the capability to write data to the project's workspace (linkfox/ directory) and perform network requests to the vendor's API gateway.
Sanitization: The retrieved patent claim text is stored and displayed without specific sanitization or filtering.
[SAFE]: The skill uses standard authentication practices. The scripts/onboarding.py utility handles user login and API key retrieval via the vendor's official domains (linkfox.com, linkfox-ai.com). Instructions for setting environment variables for these keys are presented transparently to the user.
[SAFE]: The use of a Base64-encoded string (_LOGIN_FIXED_UID) in scripts/onboarding.py is for a static JSON identifier used in the login header ({"a_id":"6a22c8b05bc916a","d_id":""}) and does not contain hidden commands or malicious payloads.

linkfox-zhihuiya-claim-data-translated