learn-from-correction
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill mentions internal scripts such as
plugins/prd-os/scripts/propose_skeptic_antipatterns.pyandplugins/kipi-core/skills/founder-voice/SKILL.md. However, it does not execute them or any arbitrary shell commands; these are listed as related references for context. - [DATA_EXFILTRATION]: The skill restricts its file system output to
q-system/output/skill-proposals/. It does not perform any network operations or access sensitive system configuration files. - [PROMPT_INJECTION]: While the skill analyzes instructions and principles, it does not contain overrides or bypasses of agent safety protocols. It explicitly emphasizes adhering to established guardrails (e.g.,
references/principle-vs-rule.md). - [SAFE]: The skill design incorporates human-in-the-loop verification by requiring that all proposed edits be manually reviewed and merged via git flow, preventing autonomous skill modification.
Audit Metadata