The Agent Skills Directory

[SAFE]: The skill implements a robust security remediation workflow that emphasizes minimal changes, confirmation of findings in the current codebase, and explicit user approval before implementation.
[SAFE]: The utility script 'scripts/parse-findings.sh' is used to parse local CSV findings and does not perform network operations or execute untrusted data.
[SAFE]: The skill presents an indirect prompt injection surface as it processes external data from findings files; however, this is a standard functional requirement for a remediation tool and is mitigated by the 'Confirm before fixing' phase and the 'hooks/hooks.json' which enforces user review before any tool-assisted edits or writes. Ingestion points: 'findings.json' and Codex CSV files. Boundary markers: None explicitly defined, but uses structured output templates. Capability inventory: High-privilege file modification tools ('Edit' and 'Write'). Sanitization: Relies on manual user review and established fix patterns.

threat-patch