canva-implement-feedback

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from external reviewers.
  • Ingestion points: The skill retrieves comment threads and replies using Canva:list-comments and Canva:list-replies in SKILL.md (Step 2).
  • Boundary markers: There are no explicit instructions to use delimiters or to treat the retrieved comment content as untrusted data, allowing potential instructions in comments to bleed into the agent's context.
  • Capability inventory: The agent has write access to the design via Canva:perform-editing-operations and can communicate externally by posting to threads via Canva:reply-to-comment.
  • Sanitization: No validation or sanitization of the comment text is performed before it is processed by the agent.
  • Reduced Oversight: The skill instructs the agent to interpret feedback "generously" and explicitly forbids asking for further confirmation after the initial plan approval ("Do NOT ask the user again"), which increases the risk that an injected instruction could be executed without specific user awareness of the malicious intent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 05:41 PM
Security Audit — agent-trust-hub — canva-implement-feedback