Babysitter plan-review-gate

Adversarial plan review by 3 independent reviewers (Feasibility, Completeness, Scope & Alignment) before presenting to user.

install

source · Clone the upstream repo

git clone https://github.com/a5c-ai/babysitter

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/a5c-ai/babysitter "$T" && mkdir -p ~/.claude/skills && cp -r "$T/library/methodologies/metaswarm/skills/plan-review-gate" ~/.claude/skills/a5c-ai-babysitter-plan-review-gate && rm -rf "$T"

manifest: library/methodologies/metaswarm/skills/plan-review-gate/SKILL.md

Plan Review Gate

Overview

Three independent adversarial reviewers validate any implementation plan before it is presented to the user. Each reviews independently with no shared context.

When to Use

After the Architect creates an implementation plan
Before presenting a plan for user approval
When validating work unit decomposition quality

Process

Feasibility Reviewer - Can this plan actually be implemented? Technical viability, resource requirements, timeline reality
Completeness Reviewer - Does the plan cover everything? Missing edge cases, untested paths, incomplete DoD items
Scope & Alignment Reviewer - Does the plan match the issue? Scope creep detection, requirement traceability

All 3 must PASS before the plan is presented.

Key Rules

Reviewers operate independently (no shared context)
Each provides a binary PASS/FAIL verdict
Findings include specific references to plan sections
Failed plans require revision before re-review

Tool Use

Invoke as part of:

methodologies/metaswarm/metaswarm-orchestrator

(Phase 1b)