The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests data from multiple project sources, including LaTeX sections and source code, which introduces a surface for indirect prompt injection. This is mitigated by the skill's restrictive scope, which focuses on pattern-based verification of numbers, symbols, and parameters rather than open-ended instruction processing. It uses specific boundary markers like decision_id for judgment claims and does not perform executable actions based on the ingested content.
[SAFE]: No malicious patterns such as credential exfiltration, unauthorized network access, or persistence mechanisms were detected. The skill's operations are confined to reading local project files and writing a structured audit report to the local file system.

consistency-auditor