assemblyai-performance-tuning
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a set of best practices for performance tuning that emphasizes security and privacy. It instructs the agent to handle the ASSEMBLYAI_API_KEY through approved secret managers and strictly forbids hardcoding or leaking credentials.
- [SAFE]: References are limited to official documentation from AssemblyAI, a well-known technology provider. No remote code execution or unauthorized package installations were detected.
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with repository content and user-provided workload profiles. It mitigates potential injection risks by instructing the agent to strictly validate output, exclude customer-derived content, and adhere to defined approval boundaries.
- Ingestion points: workload-profile argument and repository code via Read, Glob, and Grep tools as defined in SKILL.md.
- Boundary markers: Explicit instructions in the 'Approval Boundaries' and 'Output' sections of SKILL.md to ignore or exclude sensitive/external content.
- Capability inventory: Read, Glob, Grep, Write, and Edit tools for repository inspection and modification.
- Sanitization: Instructions in SKILL.md to verify the absence of sensitive fields and deterministic validation of results before reporting success.
Audit Metadata