Promptfoo

Work with Promptfoo for local, repeatable LLM evals and red-team testing. Use when a request explicitly involves Promptfoo, `promptfooconfig.yaml`, Promptfoo...

This skill appears to do what it says: help you scaffold, validate, and run Promptfoo configs. Before using it: (1) review any scaffolded promptfooconfig.yaml the script writes (especially if you pass --force) so you don’t accidentally commit secrets or point tests at sensitive APIs; (2) be aware that running promptfoo eval/validate (not included here) will contact model providers and require provider credentials you control—do not store API keys in repo files or generated prompt files; (3) confirm you install the actual Promptfoo CLI from its official source before running commands; and (4) if you plan red-team or live-API tests, avoid targeting internal or production endpoints without appropriate authorization. Overall the package is internally consistent and low-risk, but normal operational caution around provider keys and target endpoints still applies.

Type: OpenClaw Skill Name: promptfoo Version: 0.1.0 The promptfoo skill bundle provides a legitimate interface for LLM evaluation and red-teaming using the Promptfoo tool. It includes helper scripts like promptfoo-preflight.py for workspace inspection and scaffold-promptfoo-config.py for generating standard configuration templates, both of which perform safe, local-only operations. The instructions in SKILL.md and the reference documentation are well-aligned with the tool's intended use and do not contain any malicious directives or indicators of data exfiltration.