agentsop-regression-gate
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [NO_CODE]: The skill is exclusively composed of markdown documentation and metadata. It does not include any executable scripts (Python, JavaScript, shell) or compiled binaries.
- [SAFE]: The SOP describes recommended practices for LLM reliability, including human curation of golden datasets (OP-02) and version-controlled baselines (OP-06). It refers to well-known developer tools like promptfoo and LlamaIndex for implementation guidance.
- [INDIRECT_PROMPT_INJECTION]: The skill outlines a workflow for processing corpus data to generate evaluation sets. While this defines an ingestion surface, the risk is mitigated by the specified manual review process and the absence of any automated execution logic within the skill itself.
Audit Metadata