work-item

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-provided artifacts such as tasks, stories, and specifications across various modes (e.g., references/mode-refine.md, references/mode-story.md). These ingestion points lack explicit boundary markers or instructions to disregard embedded prompts within the data. Combined with capabilities like spawning sub-agents and executing scripts, this creates a vulnerability surface where a malicious artifact could influence downstream actions.
  • [DYNAMIC_EXECUTION]: The skill generates interactive HTML mockups containing embedded JavaScript for state simulation and navigation (references/mode-mockup.md, references/mockup-interactive-demo.md). These prototypes are assembled from templates and user-supplied descriptions (acceptance criteria and interaction flows). While scripted for simulation purposes only, they involve the generation and loading of executable content.
  • [COMMAND_EXECUTION]: The skill uses shell commands via Node.js to perform supporting operations, such as rendering HTML to PNG using node .claude/skills/html-export/scripts/export.cjs and opening reports using node .claude/scripts/open-report.cjs. These scripts are part of the local environment but are triggered by the skill's logic based on artifact processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 10:12 AM