project-verify-plan

Audits implementation plans for risk coverage, scope match, and edge cases before coding.

1|Updated Jun 2, 2026
One-click install
npx skills add https://github.com/envoydev/claude-stack --skill project-verify-plan-envoydev
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: project-verify-plan
Source: https://github.com/envoydev/claude-stack/tree/main/stack/skills/project-verify-plan
Command: npx skills add https://github.com/envoydev/claude-stack --skill project-verify-plan-envoydev

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? A plan built perfectly is still wrong if the plan itself was wrong. This Skill catches expensive plan defects—missing failure modes, unverified symbol names, scope drift, and silent edge cases—on the page before any code is written, where fixing them is cheapest. ## Core Features & Use Cases - Five-pass audit: Runs risk coverage, scope match, existence verification, edge/safety analysis, and soundness checks against the target stack's own trap list rather than a generic checklist. - Existence verification: Confirms every symbol, file, config key, package version, and API the plan names actually exists, flagging anything unconfirmable as unverified. - Structured verdict output: Emits seven named fields (Oriented, Asked, Decisions, Scope, Existence, Findings, Gated) plus a severity-tagged punch-list, and stamps the verdict into the plan file. - Use Case: Before building a streamed data export feature, run the audit on the plan to discover it never names cancellation handling on the stream and adds an unrequested format option—fix the plan, then build. ## Quick Start Ask the AI to review this implementation plan with project-verify-plan before I start building it.

Frequently Asked Questions about project-verify-plan

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review an implementation plan before writing code?▼

Run a structured plan audit that checks five passes in order: risk coverage against the stack's trap list, scope match against the original request, existence of every named symbol, edge and safety cases, and soundness. Fix findings in the plan before building.

What should a pre-build plan review check for?▼

Check that the plan names the stack's non-obvious failure modes, covers exactly what was asked with nothing speculative added, verifies every referenced symbol and config key exists, and calls out auth, migration, and concurrency safeguards.

When should I use plan verification versus code review?▼

Use plan verification after a plan exists but before implementation starts; it audits the design on the page. Code review happens after the build and inspects the actual diff, tests, and implementation quality.

Why verify that symbols and APIs in a plan actually exist?▼

Asserted existence is the most expensive defect class: an invented design token once cost over 2.6 million tokens to repair. One find_symbol call or config read per named thing prevents plans built on names that were never looked up.

Can the plan audit run in a separate agent instead of the current chat?▼

Yes. On request, the audit can be dispatched to the stack's verifier seat, which runs the same five passes over the plan file on its configured model and returns a punch-list. Only one seat is dispatched, with no fan-out.