evals plugins

grounded-ai/grounded_ai

Plugin Claude Code

Bundles 1 hook

Stop hook that checks Claude's final answer against the turn's tool output with TypeSafe's Jev, and sends Claude back to re-check unsupported claims.

PrefectHQ/prefect-mcp-server

Plugin Claude Code

Bundles 2 skills, 1 MCP server · 110 tokens together

Prefect workflow orchestration - hosted MCP tools for diagnostics, documentation, and release notes.

rennf93/opus-fable-playbook

Plugin Claude Code

Bundles 3 skills, 2 commands, 1 agent, 7 hooks · 227 tokens together

Make Opus 4.8 behave like Fable 5: doctrine output style, drift-catching hooks, orchestration skills, and an eval loop against golden Fable transcripts.

luanmorenommaciel/brief-spec

Plugin Claude Code

Bundles 3 skills, 5 hooks · 209 tokens together

Type-aware, evidence-backed reading and delivery for AI coding agents.

prime-radiant-inc/parallel-adversarial-review

Plugin Claude Code

Bundles 2 skills · 213 tokens together

Two skills for adversarial code review. parallel-adversarial-review dispatches two same-model reviewer subagents in parallel under a competitive scoring frame, aggregating findings and taking the worst severity on disagreement. multi-model-adversarial-review extends this to N installed coding-agent CLIs (claude, codex.

aiopshwang/goal-to-proof

Plugin Claude Code

Bundles 1 skill · 115 tokens together

Finish authorized, non-trivial AI agent work with direct, scope-matched evidence.

Abhillashjadhav/AI-PM-essential-skills

Plugin Claude Code

Bundles 5 skills, 3 agents · 899 tokens together

Five skills with manual acceptance examples that make a PM's daily Claude workflow cheaper, faster, and self-checking — now proactive by default: model-complexity-router (silently scores every task handoff, not just direct questions), builder-validator (frozen-spec self-QA, checklist frozen silently, never invents…

cyanheads/evals-mcp-server

Plugin Claude Code

Bundles 33 skills, 1 MCP server · 2,497 tokens together

Author verifiable eval records through a draft → review → revise → submit loop with server-enforced graders; compile to JSONL/CSV/Inspect/lm-eval via MCP. STDIO or Streamable HTTP.