taibu-divination
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the execution of a local Node.js script (taibu.mjs) which imports and runs a bundled computation engine (taibu-core.bundle.mjs) for various divination methods.
- [COMMAND_EXECUTION]: Instructions direct the agent to invoke the system date command to ensure time-sensitive charts are calculated using the correct current timestamp.
- [PROMPT_INJECTION]: Several tools such as liuyao and tarot accept user-supplied questions, creating a surface for potential indirect prompt injection.
- Ingestion points: The question parameter accepted by divination tools via the taibu.mjs CLI wrapper.
- Boundary markers: Instructions explicitly specify that the agent must treat the tool's calculated output as the only source of truth for interpretation, reducing the risk of the agent following instructions potentially hidden in user-provided questions.
- Capability inventory: Execution of a local Node.js wrapper script and standard system time commands.
- Sanitization: The skill utilizes a JSON-based stdin interface for the CLI tool to mitigate risks related to shell escape sequences and command injection.
Audit Metadata