anti-ui-slop-finish-gate
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted frontend source code and UI markup, creating a vulnerability surface where embedded malicious instructions could potentially influence the agent's behavior.
- Ingestion points: The skill uses
read_fileandsearch_codeto ingest external component markup and implementation details for review. - Boundary markers: The skill contains an explicit 'Security Guardrails' section with 'ASI01 Goal Hijack' instructions, directing the agent to treat reviewed markup strictly as data and ignore any instructions found in comments or data attributes.
- Capability inventory: The skill is configured with
write_file,edit_file, andrun_buildtools, which represent the capability set that could be exploited if an indirect injection were successful. - Sanitization: The skill mitigates risks by requiring the agent to follow a rigid 13-trigger checklist and a 6-state verification matrix, grounding the decision-making process in structured criteria rather than open-ended instructions.
Audit Metadata