global-skill-evolution
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local Python scripts (
scripts/skill_evolution_audit.pyandscripts/capsule_engine.py) to perform structural audits and manage a rule database. These scripts are limited to local file and database operations and do not perform network activities or utilize unsafe shell execution methods. - [INDIRECT_PROMPT_INJECTION]: The skill manages a workflow for updating instructions based on external data, creating a potential attack surface for indirect injection.
- Ingestion points: The skill ingests user corrections, failure fingerprints, and external methods during the 'discovery' phase defined in SKILL.md.
- Boundary markers: Present; the automation prompt utilizes placeholders for workspace state and the skill defines an explicit 'Output Contract' to structure incoming findings.
- Capability inventory: The skill possesses the capability to modify local instruction files and execute internal Python scripts for database management and structural auditing.
- Sanitization: Present; the skill implements rigorous 'Validation Gates', 'Change Gates', and the 'Adapted Darwin Rubric' which require evidence-driven decisions, reversible actions, and mandatory structural validation before any modifications are accepted.
Audit Metadata