cure
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data, such as CI failure summaries and reports generated by other agents, to guide automated code modifications.
- Ingestion points: Reads finding reports from the
.cheese/age/directory and accepts user-pasted findings or CI failure logs (SKILL.md). - Boundary markers: Employs a 'Coder brief' structure to scope instructions for the repair agent, including specific recommendations and invariants (
references/selection.md). - Capability inventory: The skill has the capability to modify project source code, execute local validation commands (lint, tests, builds), and dispatch publication via the
/plateskill (SKILL.md). - Sanitization: Mandates a 'Taste-test' behavioral review phase and strictly adheres to an 'Iron Law' that requires passing validation gates for all fixes before they are considered applied (
SKILL.md,references/cure-discipline.md).
Audit Metadata