The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/patent_search.py script performs network requests to OpenAlex, a well-known academic database, to retrieve bibliographic records for prior art research. This behavior is documented and targets a trusted service.
[COMMAND_EXECUTION]: The skill includes Python scripts (copyright_source_prep.py and patent_search.py) that perform local file system operations and network I/O. These scripts facilitate the primary purpose of the skill and utilize standard library modules.
[DATA_EXFILTRATION]: The skill incorporates a privacy-focused script, copyright_source_prep.py, which identifies and redacts sensitive information such as private keys, credentials, and personal contact details from source code before it is packaged for submission.
[PROMPT_INJECTION]: While the skill ingests external data (user source code and technical text), it is not a direct target of indirect prompt injection due to its focus on offline material preparation.
Ingestion points: Technical descriptions and source code processed by the provided templates and scripts.
Boundary markers: Instructions rely on standard markdown structure; no specific security boundaries are defined for processed text.
Capability inventory: File read/write access for material preparation and network read access for prior art retrieval.
Sanitization: The skill implements privacy-focused redaction for secrets and PII, although it does not specifically filter for malicious instructions within ingested code.

light-ip-application