verify

A command that performs a hostile review of the current code changes before they are declared complete. It checks the changes against a written task specification and uses a second verification pass.

In plain words
What is it for?
Use it to verify an implementation before committing or marking it done. It requires PROMPT.md or IMPLEMENTATION_PLAN.md and returns a JSON pass/fail result with failure details.
Why use it?
It looks for shortcuts, environmental problems, and other reasons a change might appear finished while still being wrong. It stops with failures instead of approving an unverified diff.

Command for Claude Code

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/archive228/loopkit/verify
Clone the repo
git clone --depth 1 https://github.com/Archive228/loopkit

Made for: Claude Code.

Per session 29 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 464 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 $0.00029 $0.00464
Opus 5 $0.00015 $0.00232
Sonnet 5 $0.00006 $0.00093
Haiku 4.5 $0.00003 $0.00046

Measured yesterday against content hash 807a10fdc8dc, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

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 yesterday.

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.

.claude/commands/verify.md · 39 lines

What it actually says

/verify — adversarial verification pass

Assume the diff is broken. Prove it isn't.

Steps

  1. Load the goal spec:

    • Read PROMPT.md if it exists. If it does not, read IMPLEMENTATION_PLAN.md.
    • If neither exists, stop and tell the user to run /spec first. Do not verify against an absent contract.
  2. Load the current diff:

    • git diff HEAD — uncommitted changes.
    • git log --oneline -5 — recent context.
  3. Invoke the adversarial-verify skill by reading skills/adversarial-verify/SKILL.md (repo layout) or .claude/skills/adversarial-verify/SKILL.md (installed layout). Walk the 11 shortcut checklist verbatim against the diff, plus the 4 environmental tells in docs/checklists/red-flags.md.

  4. Dispatch the verifier subagent (.claude/agents/verifier.md) for a second, cold-context pass on the same diff.

  5. Return a single JSON verdict, and nothing else:

    {"passes": bool, "failures": [{"file": str, "line": int, "shortcut": str, "why": str}]}
    
  6. If passes is false: do NOT commit, do NOT mark the task done, print the failure list, and stop.

Never

  • Propose fixes in this pass. Verification is separate from repair — mixing them lets the model rationalize.
  • Run application code. Read-only tools only.
  • Be polite. Politeness is how "fake done" ships.
  • Skip the shortcut checklist because "the diff is small".

Exit contract

Non-zero exit on any failure. The run.sh loop treats non-zero as "not done" and re-runs the next turn.

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. yesterday First seen · 39 lines · 29 tokens per session scan A 807a10fdc8dc

Subscribe to this mod's changes

verify is a command published in the GitHub repository Archive228/loopkit (753 stars, last pushed 1mo ago), licensed MIT. It adds 29 tokens to every session and 464 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.