preview-verdicts
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of JSON feedback pasted into the chat.
- Ingestion points: The 'Process the pasted feedback' section in SKILL.md instructs the agent to read and apply JSON data.
- Boundary markers: The skill lacks explicit instructions to ignore embedded natural language commands within the JSON fields.
- Capability inventory: The agent is instructed to modify README.md based on the input.
- Sanitization: No sanitization of the input content is performed.
- [DYNAMIC_EXECUTION]: The skill generates an executable HTML file by injecting data into a template and writing it to the local file system. This results in the creation of a local script file containing dynamically generated data arrays.
- [COMMAND_EXECUTION]: The skill uses the system 'open' command to launch the generated HTML file in a browser, which is a form of local file execution triggered by the agent.
Audit Metadata