linofcp007

EvalPlan — AI skill for Claude Code

AI community

Phase 3 (+ai) — build golden / adversarial / regression eval sets, graders, thresholds, baseline.

How to install EvalPlan

Installs to ~/.claude/skills/linofcp007-dev-spec-driven-evalplan/SKILL.md

Terminal

mkdir -p ~/.claude/skills/linofcp007-dev-spec-driven-evalplan && curl -fsSL https://raw.githubusercontent.com/linofcp007/dev-spec-driven/HEAD/commands/evalPlan.md -o ~/.claude/skills/linofcp007-dev-spec-driven-evalplan/SKILL.md

Restart Claude Code, or start a new session, for it to be picked up.

What EvalPlan does

description: Phase 3 (+ai) — build golden / adversarial / regression eval sets, graders, thresholds, baseline. PT - plano de evals (+ai). ES - plan de evals (+ai). argument-hint: "[feature name]"

Use the **dev-spec-driven** skill, Phase 3 (Eval Plan). Only relevant when the feature is on the **+ai** track.

Feature: $ARGUMENTS

Write `eval-plan.md`: a **golden** set (50–200 representative inputs with expected quality), an **adversarial** set (prompt injections, jailbreaks, out-of-scope,

Alternatives in AI

Codex — openai-codex Adversarial code review, Codex CLI integration, cross-model analysis 9.6k ★
Internal Safety Collapse — We built an adversarial codespace setup 1.2k ★
UltraCode Shim — Give Claude Code's ultracode mode to ANY model you already pay for 422 ★