red-verify

red-verify is a command for Claude Code from mrvladd-d/memobank. It costs 28 tokens per session (2,503 once invoked), scanned A, original, MIT.

A command for adversarial semantic verification: checking whether a proposed solution is actually correct in substance, not only whether it meets stated acceptance criteria.

In plain words
What is it for?
Use it to review completed work for system-level risks, incorrect assumptions, drift, and problems missed by ordinary requirement or functional checks.
Why use it?
It helps find solutions that solve the wrong problem, harm the wider system, create inconsistent state, or add hidden maintenance and operational costs.

Command for Claude Code

Written for Claude Code: $ARGUMENTS substitution. Also seen: mentions subagents.

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 commands/mrvladd-d/memobank/red-verify
Clone the repo
git clone --depth 1 https://github.com/mrvladd-d/memobank

Made for: Claude Code.

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 red-verify

README.md
[![agentmods](https://agentmods.dev/badge/commands/mrvladd-d/memobank/red-verify.svg)](https://agentmods.dev/commands/mrvladd-d/memobank/red-verify)
Your own site
<a href="https://agentmods.dev/commands/mrvladd-d/memobank/red-verify"><img src="https://agentmods.dev/badge/commands/mrvladd-d/memobank/red-verify.svg" alt="Measured on agentmods" height="20"></a>
Per session 28 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,503 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 5.1 $0.00028 $0.02503
Opus 5 $0.00014 $0.01252
Sonnet 5 $0.00006 $0.00501
Haiku 4.5 $0.00003 $0.00250

Measured 6d ago against content hash 0700ae87083c, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

red-verify 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 6d 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.

skills/_shared/references/commands/red-verify.md · 186 lines

How it starts

The opening of the file, as written. The whole thing — 186 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/red-verify — Adversarial semantic verification

Этот проход должен ловить ситуации:

  • acceptance criteria формально выполнены, но решена не та проблема
  • локально всё выглядит корректно, но решение вредит системе целиком
  • реализация переоптимизирована под локальную task interpretation и игнорирует соседние ограничения
  • появляются drift, state inconsistency, operational risks или скрытая стоимость сопровождения

Разделение ролей:

  • /verify → "выполнено ли по AC/REQ и есть ли evidence?"
  • /review → "достаточно ли качественен сам Memory Bank / planning surface?"
  • /red-verify → "это вообще хорошее и правильное решение в substance?"
  • scheduler (/autopilot / /autonomous) → task status transitions, closure, failure handling, and dependent block/unblock in scheduler mode

Status transitions have two modes.

Scheduler mode:

  • /autopilot and /autonomous own task status transitions.
  • Scheduler decides closure/failure/blocking eligibility.
  • /execute returns scoped implementation handoff; it does not close tasks.
  • /verify gives functional verdict/evidence; in scheduler mode it does not close/fail/block/promote.
  • /red-verify gives semantic verdict for T2/T3; in scheduler mode it does not close/fail/block/promote.
  • Scheduler must write the closure/failure/blocking decision, final task status, and evidence links to the authoritative indexed .memory-bank/tasks/TASK-*.task.json record before /mb-sync.
  • /mb-sync records/reconciles already-written task state. It does not decide closure/failure/blocking/promotion and must not sync a decision that exists only in scheduler context.
  • T0/T1 scheduler closure may use compact evidence / functional PASS according to tier policy.
  • T2/T3 scheduler closure requires VERDICT: PASS plus SEMANTIC_VERDICT: semantic-pass before scheduler marks done.
  • T3 scheduler closure also requires exact markers HUMAN_CHECKPOINT: done and ROLLBACK_RECOVERY_NOTE: present.

Read the full file on GitHub · 186 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. 6d ago First seen · 186 lines · 28 tokens per session scan A 0700ae87083c

Subscribe to this mod's changes

red-verify is a command published in the GitHub repository mrvladd-d/memobank (55 stars, last pushed 3mo ago), licensed MIT. It adds 28 tokens to every session and 2,503 once invoked, about $0.0001 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 2026-08-30.