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.py and plugins/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
Risk Level
SAFE
Analyzed
Jun 16, 2026, 03:09 PM
Security Audit — agent-trust-hub — learn-from-correction