jungian-psychologist
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection.
- Ingestion points: The agent is designed to ingest and interpret user-provided dream descriptions, body sensations, and emotional states. It also uses
WebFetchandfirecrawl_searchto retrieve data from external websites. - Boundary markers: The skill contains robust instructions in
SKILL.mdunder 'Ethical Boundaries' and 'Anti-Patterns' (e.g., 'not for therapy', 'explore, don't dictate') which help define the expected behavior. - Capability inventory: The agent has the ability to search the web, fetch content, and write/edit files using
Read,Write,Edit,WebFetch, andfirecrawl_searchtools. - Sanitization: While the instructions provide a strong behavioral framework, there is no technical sanitization of the input, meaning malicious instructions embedded in a user's 'dream' or an external webpage could potentially influence the agent's actions.
- [NO_CODE]: This skill contains no executable code files. All logic is implemented via natural language instructions and structured reference material in Markdown format, which eliminates risks associated with remote code execution, command injection, or malicious dependencies within the skill package itself.
Audit Metadata