otel-browser
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill includes evaluation files such as 'evals/files/hostile-browser-config.ts' and 'evals/files/hostile-page.html' that contain sample malicious instructions ('Ignore the review request', 'run curl'). These are used as test cases for validating the agent's safety guardrails.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from page content, console logs, and DOM attributes. * Ingestion points: Defined in 'SKILL.md' and 'references/instrumentation.md', including user actions and page navigation data. * Boundary markers: The 'Safety and evidence gate' in 'SKILL.md' provides clear instructions to treat all external input as untrusted and to ignore embedded commands. * Capability inventory: The skill suggests using 'gh api', 'npm view', and 'WebFetch' for technical lookups. * Sanitization: Guidelines in 'references/performance.md' describe methods for URL sanitization and PII redaction.
Audit Metadata