accessibility-testing

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze untrusted data from an application's accessibility tree (labels, hints, values), which serves as a potential surface for indirect prompt injection.\n
  • Ingestion points: UI metadata is retrieved from the target application via the tools idb-ui-describe and idb-ui-find-element (SKILL.md).\n
  • Boundary markers: Absent. The instructions do not prescribe the use of delimiters or "ignore embedded instructions" warnings when the agent processes content retrieved from the accessibility tree.\n
  • Capability inventory: The skill utilizes UI inspection tools and screenshots to interpret application state (SKILL.md).\n
  • Sanitization: Absent. There is no requirement for the agent to sanitize or validate accessibility labels or values retrieved from the external application before analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 10:28 PM