find-next-task
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill ingests untrusted data from the repository by searching for and parsing JSON plan files (docs/plans/.json, **/.json).
- Ingestion points: JSON files in the project directory are read and their content is processed.
- Boundary markers: None present. The skill does not use delimiters or instructions to prevent the agent from obeying commands embedded in task descriptions or steps.
- Capability inventory: The skill directs the agent's next steps, and since the agent likely has tools for file modification or command execution, this can lead to the execution of malicious tasks.
- Sanitization: No filtering or validation is performed on the textual content of the plans to ensure they contain only data and not instructions.
Audit Metadata