The Agent Skills Directory

[SAFE]: The skill is composed entirely of instructional markdown, reference documentation, and configuration files. It does not contain any executable code, scripts, or binaries.
[SAFE]: The references/fact-integrity.md reference provides explicit security guidelines for the agent to identify and redact sensitive information, such as API tokens (e.g., sk-live- prefixes), passwords, and private keys, if they are found in user-supplied source material.
[COMMAND_EXECUTION]: Static analysis detected a destructive command string (rm -rf /) in references/behavior-eval-fixtures.json. Contextual analysis confirms this is a negative test fixture ('expected_pass': false) used for security validation—specifically to ensure the agent rejects such commands—rather than an execution instruction.
[PROMPT_INJECTION]: The skill processes untrusted user-supplied data (notes, drafts), which constitutes an indirect prompt injection surface. However, it implements a strict 'Source Precedence' and 'Hard Rules' framework that prevents external input from overriding established safety guidelines or manufacturing claims.

human-writing