skills.rest
gmickel
gmickelCOMMUNITY
Explore all skills by this author →
Website𝕏@gmickelGitHub

flow-next-spec-completion-review

Verifies that completed spec tasks fully implement all spec requirements before close.

692|55|Updated Dec 26, 2025
Search Tags:#spec review#multi-backend#receipt#Flow-Next#requirements verification#Code review#Workflow Automation
One-click install
npx skills add https://github.com/gmickel/gmickel-claude-marketplace --skill flow-next-spec-completion-review
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: flow-next-spec-completion-review
Source: https://github.com/gmickel/gmickel-claude-marketplace/tree/main/plugins/flow-next/codex/skills/flow-next-spec-completion-review
Command: npx skills add https://github.com/gmickel/gmickel-claude-marketplace --skill flow-next-spec-completion-review

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

When an agent finishes all tasks in a spec, there is no guarantee the combined implementation actually covers every requirement — decomposition gaps, partial implementations, and scope drift slip through. This Skill runs an independent completion review that checks spec compliance (not code quality) before a spec is allowed to close.

Core Features & Use Cases

  • Multi-backend review dispatch: Routes the review to Codex CLI, GitHub Copilot CLI, Cursor CLI, RepoPrompt, or a host-native cross-family subagent, resolved via flag, env var, or config.
  • Requirement coverage verification: Extracts every spec requirement and R-ID, maps them to the diff, and emits SHIP / NEEDS_WORK / NEEDS_HUMAN verdicts with receipts.
  • Bounded fix loop with durable state: Automatically fixes findings and re-reviews up to a round cap, persisting terminal status, receipts, and review-round counters so interrupted runs resume safely.
  • Use Case: After all tasks of spec fn-22 are marked done, run the completion review to confirm every acceptance criterion is implemented; on NEEDS_WORK it fixes gaps and re-reviews until SHIP or escalation.

Quick Start

Ask the agent to run the flow-next spec completion review for spec fn-22 before closing it.

Frequently Asked Questions about flow-next-spec-completion-review

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

FAQPage Schema
How do I run a spec completion review in flow-next?▼

Invoke the skill with a spec id, for example `fn-22`, optionally adding `--review=codex|copilot|cursor|host|rp|none` to override the backend. The backend is otherwise resolved from FLOW_REVIEW_BACKEND or .flow/config.json via `flowctl review-backend`.

What is the difference between spec completion review and impl-review?▼

Completion review checks spec compliance only: whether every requirement and R-ID is implemented across all tasks. Code quality, style, and test quality per task are handled by impl-review, which completion review explicitly does not duplicate.

Which review backends does flow-next completion review support?▼

It supports Codex CLI, GitHub Copilot CLI, Cursor CLI (cursor-agent), RepoPrompt on macOS, and a host-native fresh subagent pinned to a cross-family model. Backend selection priority is the --review flag, then FLOW_REVIEW_BACKEND, then .flow/config.json.

What happens when the completion review returns NEEDS_WORK?▼

The skill parses the findings, fixes the code, runs tests, commits, and re-reviews in an internal loop until SHIP or the round cap (default 8). At the cap it writes needs_work status and escalates with exit 4 instead of looping forever.

Why does the host backend require a cross-family model pin?▼

The host reviewer must come from a different model family than the one that wrote the implementation to avoid self-review bias. If no cross-family pin exists in AGENTS.md routing, the review fails closed or escalates to NEEDS_HUMAN in autonomous mode.

Can the completion review be skipped?▼

Yes, passing `--review=none` skips the review after informing the user, and a work 3g policy skip persists `not_required` status. However, review is never skipped silently, and gates requiring completion review will keep demanding it while status is unknown.

Related Skills

Explore More Software Engineering Skills→
tiendungtiendung

flow-next-plan

Convert feature requests or Flow IDs into epics and tasks in .flow/ via flowctl.

Community
Advanced
uk-agentsuk-agents
9

gaps

Report open compliance gaps and update tracker statuses via close or accept modes.

Official
Intermediate
silcamsilcam

quality-review

Review pull request diffs for code quality and test coverage.

Official
Advanced
knowledgeislandsknowledgeislands

ki-agents

Audit Claude Code subagent definitions against mechanical and judgment criteria.

Official
Advanced