auditing-experiments-flags
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs configuration audits using internal platform tools. It evaluates metadata for experiments and feature flags to identify hygiene and process issues. All data access is handled through authorized tools (e.g.,
experiment-get,feature-flag-get-all) and results are presented to the user through standard markdown reports or internal notebook creation.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from PostHog entities, which presents a surface for indirect prompt injection.\n - Ingestion points: Metadata such as descriptions and hypotheses are fetched via
experiment-getandfeature-flag-get-definition(referenced inSKILL.mdandreferences/experiment-checks.md).\n - Boundary markers: The instructions do not specify the use of delimiters or boundary markers when processing these fields.\n
- Capability inventory: The skill's actions are restricted to reading data and reporting findings; it does not possess capabilities for arbitrary command execution, network exfiltration, or filesystem writes outside of the notebook environment.\n
- Sanitization: No specific sanitization or filtering of external content is mandated before interpolation into reports.
Audit Metadata