figma-handoff-gate
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Figma text nodes, frame names, and descriptions, creating a surface for embedded instructions to influence the agent.
- Ingestion points:
SKILL.md(Checks 11 and 16) reads content fromn.characters,n.name, andn.descriptionacross various Figma nodes. - Boundary markers: Absent; the skill does not implement delimiters or warnings to ignore potentially malicious instructions inside the ingested data.
- Capability inventory: The agent uses
figma_executefor multiple auditing tasks and has access tofigma_capture_screenshotfor visual proof. - Sanitization: Absent; external content from the Figma document is used directly in logic (regex, list matching) without escaping or validation.
- [DYNAMIC_EXECUTION]: The skill requires the agent to generate and execute code at runtime to resolve audit findings.
- Evidence: The "Apply-fix commands" section in
SKILL.mdinstructs the agent to dynamically "write that code for the approved fix" and execute it via thefigma_executetool. - [COMMAND_EXECUTION]: The skill provides predefined JavaScript snippets to be executed against the Figma Plugin API.
- Evidence: Several checks (e.g., Check 11, 15, 16, and 17b) provide JavaScript blocks intended to be run through the
figma_executetool, allowing the agent to read and modify the active Figma document environment.
Audit Metadata