wcag-keyboard-navigation
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes multiple evaluation tasks that require an agent to ingest and analyze untrusted data, such as accessibility audit reports and user-provided feedback.
- Ingestion points: The
task.mdfiles in theevals/subdirectories (e.g.,evals/aa-signoff-on-a-checkout-with-two-sticky-bars/task.md) process data from files likereports/keyboard-sweep.md,docs/a11y-backlog.md, andaudit/finding-17.md. - Boundary markers: The tasks use standard markdown structure and code blocks but lack explicit instructions to the agent to treat embedded data as untrusted or to ignore instructions contained within it.
- Capability inventory: The skill enables and requires the agent to perform file modifications (
src/*.js,src/*.css) and create documentation based on the provided inputs. - Sanitization: No sanitization or validation logic is present for the input data processed during these evaluation tasks.
Audit Metadata