AGENT_RUN_VERDICT_WORKFLOW

A workflow for making each in-app agent run end with an operator-facing verdict. The verdict is backed by stored evidence and citations, shown in the existing interface, and paired with explicit next actions.

In plain words
What is it for?
Use it when adding or reviewing agent-run contracts, issue context, live workflow panels, verdict displays, citations, trace evidence, and deterministic tests.
Why use it?
It keeps backend data, agent results, interface panels, evidence, tests, and verification aligned instead of treating them as separate pieces.

Agent

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/homenshum/nodebenchai/agent_run_verdict_workflow
Clone the repo
git clone --depth 1 https://github.com/HomenShum/NodeBenchAI
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,525 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. 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 $0.00000 $0.01525
Opus 5 $0.00000 $0.00763
Sonnet 5 $0.00000 $0.00305
Haiku 4.5 $0.00000 $0.00153

Measured 2d ago against content hash ee73b446d3b6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

AGENT_RUN_VERDICT_WORKFLOW 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.

docs/agents/AGENT_RUN_VERDICT_WORKFLOW.md · 205 lines

How it starts

The opening of the file, as written. The whole thing — 205 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Run Verdict Workflow

Goal

Make every in-app agent run capable of producing a final operator-facing verdict with open-source citations, trace-backed evidence, and explicit next actions, all surfaced through the existing UI.

This is the standard workflow for any agent-facing slice that touches:

  • contract + data model wiring
  • backend issue context enrichment
  • frontend live workflow panels
  • verdict exactness UI surfacing
  • tests + verification

This extends the existing NodeBench harness. Do not build a second orchestration stack, second trace system, or parallel review UI.

Operating stance

  • Direct yourself as if you were the product owner and reliability lead.
  • Do not stop after a backend patch if the UI cannot expose the result.
  • Do not stop after UI polish if the verdict is not defensible from stored evidence.
  • Do not stop after a “looks good” pass if deterministic tests and verification are missing.

Workflow contract

1. Contract + data model wiring

Start at the contract layer.

Define or confirm:

  • the user-facing job contract
  • the stored data contract
  • the verdict contract
  • the UI contract

For agent-task substrate work, prefer the existing tables and query surfaces:

  • agentTaskSessions
  • agentTaskTraces
  • agentTaskSpans
  • toolApprovals
  • existing sourceRefs, goalId, visionSnapshot, successCriteria, crossCheckStatus, deltaFromVision

If the feature needs new session output, first ask:

  1. Can this be derived from existing session or trace metadata?
  2. Can the UI query synthesize it without adding new persistence?
  3. If persistence is required, is the new field part of the canonical session contract?

Rules:

  • Prefer derived proof packs over duplicate storage.
  • Keep verdict enums explicit and bounded.
  • Keep source references structured, not embedded in prose only.
  • Use exact validators and exact return contracts in Convex.

Minimum verdict shape:

  • verdict
  • verdictLabel
  • summary
  • confidence
  • evidenceCount
  • citationCount
  • verificationCounts
  • approvalCounts
  • keyFindings
  • openIssues
  • nextActions
  • topSourceRefs

Read the full file on GitHub · 205 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 First seen · 205 lines · 0 tokens per session scan A ee73b446d3b6

Subscribe to this mod's changes

AGENT_RUN_VERDICT_WORKFLOW is an agent published in the GitHub repository HomenShum/NodeBenchAI (14 stars, last pushed 19d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,525 tokens. 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 2026-08-30.