signals-scout-replay-vision
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns or security risks were identified. The skill is a monitoring tool for PostHog Replay Vision, authored by the vendor, and follows standard operational patterns using authorized tools.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and proactively mitigates the risk of indirect prompt injection originating from untrusted LLM-generated scanner outputs.
- Ingestion points: SQL queries against the
$recording_observedevent properties in the PostHog events table. - Boundary markers: The prompt includes explicit instructions to treat scanner-derived text as untrusted data and to ignore any potential instructions or commands embedded within it.
- Capability inventory: SQL execution via
execute-sql, state persistence viascout-scratchpadtools, and reporting viascout-emit-report. - Sanitization: The instructions mandate the use of sanitized identifiers for memory keys and the truncation of untrusted prose strings to prevent processing large or malicious payloads as logic.
Audit Metadata