form-palette
Pass
Audited by Gen Agent Trust Hub on Jul 1, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The 'Delivery' section contains instructions that attempt to control and potentially suppress agent reporting behavior. It directs the agent to 'Never dump analysis to CLI' and instead use a specific tool ('/atlas-report') with a format including 'severity indicators' and a 'verdict'. This could be used to hide the agent's internal reasoning or technical output from the user.
- [COMMAND_EXECUTION]: The skill workflow interpolates the user-provided '{product_type}' variable directly into shell commands ('python3 -m form_agent.uiux search') without visible sanitization or validation, creating a surface for indirect command injection.
- Ingestion points: The '{product_type}' variable is sourced from user requests or project context (SKILL.md).
- Boundary markers: No delimiters or 'ignore' instructions are used around the variable in the bash commands.
- Capability inventory: The skill explicitly requests the 'Bash' tool to execute vendor-specific Python modules.
- Sanitization: There are no instructions for the agent to sanitize, escape, or validate the user-supplied string before execution.
- [PROMPT_INJECTION]: There is a version mismatch between 'SKILL.md' (0.6.6) and 'plugin.json' (1.9.1), which represents deceptive or inconsistent metadata.
Audit Metadata