The Agent Skills Directory

[COMMAND_EXECUTION]: Executes bundled Python scripts (fiscal_year_calculator.py, mechanism_matcher.py, citation_tracker.py) to provide deterministic calculations for fiscal year windows and grant mechanism eligibility. It also utilizes the bash_tool to interact with the official NIH RePORTER API using curl POST requests for grant discovery.
[EXTERNAL_DOWNLOADS]: Communicates with official government domains (api.reporter.nih.gov, grants.nih.gov) and the Consensus research platform (consensus.app) to retrieve project data and research literature. These connections are strictly scoped to the skill's primary research purpose.
[PROMPT_INJECTION]: The skill contains 'Agent Integrity Rules' designed to prevent the AI from hallucinating data or using prior training knowledge in place of session-retrieved facts. This effectively mitigates certain types of instruction drift and ensures data provenance.
[DATA_EXFILTRATION]: No unauthorized data exfiltration patterns were detected. Network activities are restricted to querying trusted research repositories and federal grant databases. The skill maintains a local session log at ~/.grants_sessions/ to track and audit data counts.

grants