goal-evaluator-judge

goal-evaluator-judge is an agent for Claude Code from synaptiai/synapti-marketplace. It costs 161 tokens per session (3,167 once invoked), scanned A, original, Apache-2.0.

An independent judge for deciding whether a software task meets its stated acceptance criteria, using test results, build logs, and other evidence.

In plain words
What is it for?
Use it to return a PASS, FAIL, or human-review-needed verdict for each acceptance criterion.
Why use it?
It separates verification from the person or agent that wrote the code, helping avoid treating implementation reasoning as proof that the result works.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter.

Part of the flow plugin — 32 skills, 21 commands, 9 agents shipped together

Good fit Use it to return a PASS, FAIL, or human-review-needed verdict for each acceptance criterion.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/synaptiai/synapti-marketplace/goal-evaluator-judge
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.

Clone the repo
git clone --depth 1 https://github.com/synaptiai/synapti-marketplace

Made for: Claude Code.

Or install flow, the plugin that ships this one along with the rest of its 32 skills, 21 commands, 9 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for goal-evaluator-judge

README.md
[![agentmods](https://agentmods.dev/badge/agents/synaptiai/synapti-marketplace/goal-evaluator-judge/github.svg)](https://agentmods.dev/agents/synaptiai/synapti-marketplace/goal-evaluator-judge)
Your own site
<a href="https://agentmods.dev/agents/synaptiai/synapti-marketplace/goal-evaluator-judge"><img src="https://agentmods.dev/badge/agents/synaptiai/synapti-marketplace/goal-evaluator-judge/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for goal-evaluator-judge

Your own site · 80×15
<a href="https://agentmods.dev/agents/synaptiai/synapti-marketplace/goal-evaluator-judge"><img src="https://agentmods.dev/badge/agents/synaptiai/synapti-marketplace/goal-evaluator-judge.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 161 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,167 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00161 $0.03167
Opus 5 $0.00081 $0.01584
Sonnet 5 $0.00032 $0.00633
Haiku 4.5 $0.00016 $0.00317

Measured 2d ago against content hash 556d803aa489, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

goal-evaluator-judge scanned grade A with 0 findings 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 2d 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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/flow/agents/goal-evaluator-judge.md · 168 lines

How it starts

The opening of the file, as written. The whole thing — 168 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Goal Evaluator Judge Agent

You are an independent loop-time verdict judge for the flow plugin's goal evaluator. Your role is to evaluate whether a FlowGoal has been satisfied based solely on file-backed evidence — never on code-writing rationale, diffs, transcript prose, or planning decisions.

Relationship to verdict-judge

You are a specialization, not a replacement:

Aspect verdict-judge goal-evaluator-judge (this agent)
Invocation One-shot per PR (Phase 4 of /flow:start) Repeated per turn (Stop hook or /flow:goal evaluate)
Verdict shape PASS | FAIL | NEEDS-HUMAN-REVIEW achieved | not_achieved | blocked | needs_human_review
Output fields per-criterion table per-criterion table + confidence + delta + next_step_hint
Time horizon terminal (gate before merge) iterative (continues the loop)
Surface Phase 4 verdict comment on PR Stop-hook decision:block/approve JSON

When in doubt about ANY rule of judgment, defer to verdict-judge's Independence Protocol. This agent inherits it verbatim and adds loop-specific fields.

Independence Protocol (inherited from verdict-judge, mechanically enforced)

You MUST NOT have access to:

  • The code diff
  • The decision journal
  • Planning notes or task decomposition rationale
  • Self-review findings from the code-writing agent
  • Project memory from previous sessions
  • Test source files — test inputs and expected values reach you only as rows inside evidence sidecars and their embedded raw output; you never open a test file to "check"
  • Screenshot files — the screen reaches you only as the analysis text inside a visual_result sidecar; a screenshot path is a pointer for the human reader, never something you open
  • The conversation transcript (Stop hook never reads it; the assembler never embeds it)

You ONLY receive (delivered inline in the prompt; you do NOT navigate the filesystem):

  1. <<<UNTRUSTED_GOAL_CONTRACT>>> — the FlowGoal YAML (outcome + acceptance criteria + verification commands + per-AC must_pass flags)
  2. <<<UNTRUSTED_DETERMINISTIC_REPORT>>> — the JSON report from flow-run-deterministic-checks.sh (per-AC pass/fail/incomplete + path violations)
  3. <<<UNTRUSTED_EVIDENCE_LEDGER>>> — every .flow/runs/<run-id>/evidence/*.evidence.yaml sidecar, concatenated, each followed by its raw output (truncated to 8KB)
  4. <<<UNTRUSTED_PREVIOUS_VERDICT>>> — last turn's verdict JSON (for delta computation: did the pass-set move forward or stay stuck?) — absent on the first turn
  5. <<<UNTRUSTED_BUDGET>>>lifecycle.turns_evaluated, continuation.max_iterations, remaining

Read the full file on GitHub · 168 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. 2d ago Changed · +2 lines 556d803aa489
  2. 8d ago First seen · 166 lines · 161 tokens per session scan A 88547779f274

Subscribe to this mod's changes

goal-evaluator-judge is an agent published in the GitHub repository synaptiai/synapti-marketplace (6 stars, last pushed yesterday), licensed Apache-2.0. It adds 161 tokens to every session and 3,167 once invoked, about $0.0008 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other agents, from other repositories

qa-tester

The QA Tester of the aSPARK team. Use in the Review phase (/demo-day) to test the running application hands-on in a real browser: verify every acceptance criterion from the spec, explore beyond the happy path, check console and network, and file reproducible bugs. Requires a browser integration (Claude in Chrome…

a-lottes/aSPARK · 79 tokens

qa-engineer

QA engineering specialist. Invoke for test plans, test case design, edge case identification, test automation strategy, quality standards, exploratory testing guidance, and defect analysis. Returns structured test artefacts and quality assessments grounded in risk-based thinking.

pranav8494/team-of-agents · 50 tokens

loop-verifier

Runs the project's verification contract against real behaviour and judges honestly - verified, gap, or failed. Does not implement fixes.

cbdreamer11/CB-loop-kit-claude-plugin · 28 tokens

browser-controller

Headless browser control via Playwright for QA testing, screenshots, form interaction, page snapshots, and accessibility auditing. The eyes of ProductionOS.

ShaheerKhawaja/ProductionOS · 32 tokens

maintainer-test-adequacy-reviewer

Use this agent before signing a maintainer Build gate on the sdlc-plugin repo. Reviews the current diff and the touched test files for test-adequacy concerns: coverage of changed lines, missing edge-case and failure-path tests, integration vs unit balance, no test-only mocks that mask production behavior. Spawned in…

lantisprime/claude-sdlc · 0 tokens

rot-claude

The 🧭 Claude lens — Praxis, empirical verification, craft. Reality is the judge. Summon when a claim needs to become a measurement — run it, build it, read it, cite it file:line — when a plan must survive contact with the real system, when "should work" must become "ran, exit 0", or when the craft gate must decide…

Nova-Violet-Role/RoT-MoE · 92 tokens