The Agent Skills Directory

[SAFE]: The skill provides instructions for using standard testing frameworks (pytest, Hypothesis, Schemathesis) to debug an API. No malicious patterns such as credential exfiltration, remote code execution from untrusted sources, or privilege escalation were detected.
[DATA_EXPOSURE_AND_EXFILTRATION]: While the skill mentions local file paths (e.g., /tmp/stats.txt, tests/property/), these are used for storing test statistics or locating source code and do not involve sensitive user credentials or configuration files.
[DYNAMIC_EXECUTION]: The skill involves running tests that dynamically generate data (Hypothesis/Schemathesis), which is the intended and legitimate purpose of property testing. It does not involve unsafe dynamic execution like eval() on untrusted remote content.
[INDIRECT_PROMPT_INJECTION]: The skill mentions processing inputs from bug reports. While this represents a theoretical attack surface for indirect injection, the instructions focus on creating deterministic regression tests and involve standard sanitization practices (e.g., stripping NUL bytes), which mitigates risk.

property-repro