The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were detected across the skill's instructions, documentation, or scripts. The skill functions as intended for security auditing.
[PROMPT_INJECTION]: The skill does not contain instructions that attempt to override agent behavior, bypass safety filters, or extract system prompts.
[DATA_EXFILTRATION]: No network operations or hardcoded credentials were found. The skill configuration explicitly disallows the curl tool, and the analysis script scan_eval.py does not perform any external network requests.
[REMOTE_CODE_EXECUTION]: The skill performs static regex-based scanning of source code. It does not execute the code it scans, and it does not download or execute remote scripts.
[INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface as it reads external source trees. However, it treats the content as data for regex matching and does not interpret instructions embedded within the scanned files. The deterministic static detector finding in references/THEORY.md is a false positive, as it correctly identifies an educational example of insecure code used for demonstration purposes.

detecting-eval-exec-usage