bug-screenshot-annotation
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill manages several local processes, including starting a Python HTTP server with
python3 -m http.server, utilizing a browser-automation CLI (playwright-cli) for rendering, and executing a TypeScript script viabunto interact with Jira. - [DYNAMIC_EXECUTION]: The workflow involves generating Python snippets for image cropping and building dynamic HTML/CSS files for visual overlays. These locally-generated artifacts are executed at runtime to produce the final annotated image.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from bug screenshots and accessibility snapshots, which could contain malicious text or structural patterns designed to influence the agent.
- Ingestion points: External screenshot files and accessibility metadata (SKILL.md, Step 1).
- Boundary markers: Absent; there are no specific delimiters or instructions to the agent to ignore embedded content within the screenshot data.
- Capability inventory: File system access, local code execution (Python/HTML), browser automation, and Jira API interaction (via bun script).
- Sanitization: None; the skill does not validate or sanitize the content of the image or its metadata before processing.
Audit Metadata