The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/verify_no_fabrication.py) to perform its primary safety function. This execution is limited to internal logic and test suites, ensuring that factual claims are verified against the user's source material.
[EXTERNAL_DOWNLOADS]: The skill frontmatter permits the use of web search and fetch tools to gather company context. The instructions contain explicit safeguards ensuring that data retrieved from the web is treated as untrusted and is not used to automatically fabricate factual claims about the user.
[SAFE]: The skill implements a robust human-in-the-loop safety model. It uses a dedicated verification script to flag new proper nouns, numbers, and phrases for user review, effectively mitigating risks associated with AI hallucinations and providing protection against potential indirect prompt injection from job postings.

cover-letter