axiom-validate-screenshots
Pass
Audited by Gen Agent Trust Hub on Jun 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a bash command loop to execute the
sipsutility on files discovered via the Glob tool. This involves the dynamic construction and execution of shell commands on the local filesystem to retrieve image dimensions. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it uses vision tools to analyze content from external image files. Malicious instructions rendered within a screenshot could potentially influence the agent's behavior.
- Ingestion points: The agent reads and analyzes image files from a user-specified folder using the
Readtool as described inSKILL.md(Step 4). - Boundary markers: There are no explicit boundary markers or instructions provided to the vision model to ignore or sequester text found within the screenshots.
- Capability inventory: The skill has access to
bash(used forsipsin Step 3) andGlob(used for file discovery in Step 2). - Sanitization: No sanitization or content filtering is applied to the visual data or any OCR-derived text before it is processed by the agent.
Audit Metadata