self-motivation-classical
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill provides an execution framework where the agent processes user tasks and 'failure contexts' while following a perfectionistic persona. It encourages active tool usage for searching and verification but fails to define boundary markers or explicit safety instructions for handling untrusted data. This establishes a vulnerability surface for indirect prompt injection, where malicious instructions embedded in processed user data could influence the agent's actions.
- Ingestion points: User tasks and 'failure context' contexts used in the seven-item checklist (found in SKILL.v2.md).
- Boundary markers: No delimiters or isolation markers are present to distinguish internal skill instructions from external data.
- Capability inventory: The skill explicitly instructs the agent to 'use tools to search' and 'verify assumptions' (found in SKILL.v2.md).
- Sanitization: No sanitization, escaping, or validation of user-provided content is defined in either file.
Audit Metadata