The Agent Skills Directory

[PROMPT_INJECTION]: Static detectors flagged prompt injection patterns in 'references/ai-agent-security.md' and 'references/owasp-checklists.md'. A manual review confirms these are educational examples and attack signatures provided as reference material for the security auditor role. They do not attempt to override the agent's system instructions or bypass safety filters.
[EXTERNAL_DOWNLOADS]: The skill's scripts are designed to work entirely with the Python standard library (stdlib). The 'scripts/requirements.txt' file explicitly states that no external dependencies are required, reducing supply chain risk. No automated downloads or execution of remote code were found.
[DATA_EXFILTRATION]: No network operations or data exfiltration patterns were identified. The scripts operate locally, reading files from a user-provided target directory and saving reports to a local 'data/reports/' folder.
[COMMAND_EXECUTION]: The skill provides several Python scripts (e.g., 'full_audit.py', 'quick_scan.py') meant to be run by the user. These scripts perform static analysis using regular expressions and do not execute the content of the files being audited.
[INDIRECT_PROMPT_INJECTION]: As a security auditor, the skill is designed to ingest and process untrusted data (source code). While this constitutes an attack surface, the scripts do not evaluate or execute the processed data as code, which significantly mitigates the risk of indirect injection attacks.

007