05-character-identity-onboarding

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill instructions and associated scripts do not exhibit any malicious patterns. The logic is focused on data structure consistency and consent recording for AI character representations.
  • [COMMAND_EXECUTION]: The eval/run-eval.sh script executes local Python scripts (contract_check.py and an inline Python snippet) to validate that identity records conform to the expected schema. This is a standard automated testing procedure and does not involve untrusted input.
  • [CREDENTIALS_SAFE]: The test fixtures in eval/fixtures/normalized-representations.json mention USEAPI_NET_TOKEN and HEYGEN_API_KEY within descriptive note fields. These are used as documentation placeholders and do not contain actual sensitive credentials.
  • [INDIRECT_PROMPT_INJECTION]: While the skill involves processing and creating markdown files (characters/<slug>.md) that may be read by other agents, it implements clear data boundaries and validation checks via the contract_check.py script.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 12:35 PM
Security Audit — agent-trust-hub — 05-character-identity-onboarding