experiment-orchestration
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes untrusted external data.
- Ingestion points: The skill reads data from the target repository/codebase in Step 2 and retrieves metric metadata (keys and descriptions) via the
get-metricstool in Step 1. - Boundary markers: The instructions do not define explicit boundary markers (such as XML tags or specific delimiters) or instructions to ignore embedded commands when interpolating this external content into prompts for the coding agent.
- Capability inventory: The skill has significant write capabilities, including the ability to modify feature flags, create metrics, and manage experiment iterations through the LaunchDarkly MCP server, as well as proposing code changes via GitHub.
- Sanitization: There is no evidence of sanitization or validation logic for the content retrieved from the codebase or metric definitions before it is used to influence the agent's logic or hypothesis generation.
Audit Metadata