verification-agent

A coding agent that checks whether completed software work meets its requirements and acceptance criteria. Acceptance criteria are the specific conditions a task must satisfy.

In plain words
What is it for?
Use it to verify completed tasks against a product requirements document, test user flows and edge cases, compare pages with Figma designs, and run Lighthouse audits.
Why use it?
It finds missing behavior, visual differences, text or translation problems, accessibility issues, performance concerns, and other defects before work is accepted.

Agent

Part of the belmont plugin — 17 skills, 6 agents shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/blake-simpson/belmont/verification-agent
Clone the repo
git clone --depth 1 https://github.com/blake-simpson/belmont

Or install belmont, the plugin that ships this one along with the rest of its 17 skills, 6 agents.

Per session 2 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 5,237 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00002 $0.05237
Opus 5 $0.00001 $0.02618
Sonnet 5 $0.00000 $0.01047
Haiku 4.5 $0.00000 $0.00524

Measured 3d ago against content hash 049e1b3368b1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

verification-agent scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

- Poll for readiness: `curl -s -o /dev/null -w "%{http_code}" $BELMONT_BASE_URL/` (or `http://localhost:$FREE_PORT/` for secondary servers) in a loop, max 60s.
plugin/agents/verification-agent.md · 367 lines

How it starts

The opening of the file, as written. The whole thing — 367 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Belmont: Verification Agent

You are the Verification Agent. Your role is to verify that task implementations meet all requirements from the PRD and acceptance criteria. You run in parallel with the Code Review Agent.

Core Responsibilities

  1. Verify Acceptance Criteria - Check each criterion is satisfied
  2. Visual Verification - Compare implementation to Figma designs using Playwright headless
  3. Check i18n/Text - Verify all text uses proper i18n keys
  4. Functional Testing - Test happy paths, edge cases, accessibility
  5. Report Issues - Document any problems found
  6. Lighthouse Audit - Run performance, accessibility, best practices, and SEO audits on public pages
  7. Cleanup - Remove all temporary verification artifacts (screenshots, reports)

Input: What You Read

You will receive a list of completed tasks and file paths in the sub-agent prompt. Tasks to verify are those marked [x] (done, not yet verified) in PROGRESS.md. Additionally, read:

  • The PRD file (at the path specified in the orchestrator's prompt) - Task definitions and acceptance criteria (pure spec, no status markers)
  • The PROGRESS file (at the path specified in the orchestrator's prompt) - Task states: [ ] todo, [>] in_progress, [x] done (not verified), [v] verified, [!] blocked, [-] withdrawn (deliberately dropped — not work, do not verify it). Letters are case-insensitive.
  • The TECH_PLAN file (at the path specified in the orchestrator's prompt, if it exists) - Technical specifications and verification requirements
  • Archived MILESTONE files (in the same directory as the PRD, matching MILESTONE-*.done.md) - Implementation context from previous phases, including design specifications, codebase analysis, and implementation logs

State updates: On verification pass, the orchestrator marks tasks [v] in PROGRESS.md. On verification fail, the orchestrator adds new [ ] follow-up tasks. You do NOT update state files yourself — only report results.

Read the full file on GitHub · 367 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 367 lines · 2 tokens per session scan A 049e1b3368b1

Subscribe to this mod's changes

verification-agent is an agent published in the GitHub repository blake-simpson/belmont (4 stars, last pushed 14d ago), licensed Apache-2.0. It adds 2 tokens to every session and 5,237 once invoked, about $0.0000 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories

nw-acceptance-designer

Use for DISTILL wave — designs E2E acceptance tests from user stories and architecture using Given-When-Then format. EXPANDED scope (plan v3 §3.A, 2026-05-19) — exclusive test-expertise owner; authors ATs with maximum PBT + parametrize density, runs self-completeness audit (7-category taxonomy + 15-item checklist)…

nWave-ai/nWave · 0 tokens

nw-nwave-buddy

Use for any nWave question — methodology, project navigation, command help, wave status, migration, and troubleshooting. The first agent to consult when unsure about anything in nWave.

nWave-ai/nWave · 0 tokens

nw-product-owner-reviewer

Use as hard gate before DESIGN wave - validates journey coherence, emotional arc quality, shared artifact tracking, Definition of Ready checklist, LeanUX antipatterns, and story sizing. Blocks handoff if any critical issue or DoR item fails. Runs on Haiku for cost efficiency.

nWave-ai/nWave · 0 tokens

nw-researcher

Use for evidence-driven research with source verification. Gathers knowledge from web and files, cross-references across multiple sources, and produces cited research documents.

nWave-ai/nWave · 0 tokens

nw-documentarist

Use for documentation quality enforcement using DIVIO/Diataxis principles. Classifies documentation type, validates against type-specific criteria, detects collapse patterns, and provides actionable improvement guidance.

nWave-ai/nWave · 0 tokens

nw-troubleshooter

Use for investigating system failures, recurring issues, unexpected behaviors, or complex bugs requiring systematic root cause analysis with evidence-based investigation.

nWave-ai/nWave · 0 tokens