Source: Promptfoo docs: Red team plugins and quickstart
Promptfoo's catalogue of 157 red-team plugins (PII, BOLA/BFLA, RBAC, SSRF, injection, MCP, cross-session leak) and the setup/run/report quickstart flow.
Contents
- What: official tool documentation (plugins reference, quickstart, configuration)
- Publisher: Promptfoo (acquisition by OpenAI announced March 2026)
- Date: accessed 2026-11-14
- Original: Plugins · Quickstart · Configuration
Summary
Complements the red team overview source with the plugin catalogue and the CLI workflow. Plugins are the “what to test” unit; each generates adversarial cases tailored to the declared application purpose.
Key claims
- 157 plugins in six categories: Brand, Compliance and Legal, Dataset, Security and Access Control, Trust and Safety, Custom.
- Privacy plugins:
pii:direct,pii:session,pii:social,pii:api-db. - Security and access-control plugins include
bola,bfla,rbac,ssrf,sql-injection,shell-injection,tool-discovery,indirect-prompt-injection,prompt-extraction,cross-session-leak,hijacking,mcp;excessive-agencysits under Brand. Several (marked in the docs) use remote generation in the Community edition. - CLI flow:
promptfoo redteam setup(web UI) orredteam init --no-gui→redteam generate(writesredteam.yaml) →redteam run(generate plus eval, keeping tests in sync with config) →redteam report. - More config keys:
maxConcurrency(default 4),delay,frameworks,testGenerationInstructions,graderExamples, per-plugingraderGuidance; the attackerproviderdefaults toopenai:gpt-5and also grades. - The
frameworksfilter acceptsowasp:llm,owasp:api,owasp:agentic,mitre:atlas,nist:ai:measure,eu:ai-act,iso:42001,gdpr,dod:ai:ethics. - Strategies listed include
basic,jailbreak,jailbreak:meta,jailbreak:composite, tree-based (TAP),crescendo,goat,best-of-n,likert,citation,math-prompt,layer,retry, and encodings (base64,rot13,leetspeak,hex,homoglyph,morse-code,pig-latin,camelCase).