verify

verify is a skill for Claude Code from angad-kandhari/deliberate. It costs 69 tokens per session (1,349 once invoked), scanned A, original, Apache-2.0.

A checklist for coding agents to prove that a change really works before calling it finished. It covers running the relevant code and checks, including behavior that tests may miss.

In plain words
What is it for?
Use it before claiming a task is done, reporting progress, or handing a change to someone else.
Why use it?
It reduces the risk of reporting success when code only looks correct or when a test was weakened to pass.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the deliberate plugin — 11 skills, 1 hook shipped together

Good fit Use it before claiming a task is done, reporting progress, or handing a change to someone else.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/angad-kandhari/deliberate/verify
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.

Any agent
npx skills add angad-kandhari/deliberate --skill verify
Clone the repo
git clone --depth 1 https://github.com/angad-kandhari/deliberate

Made for: Claude Code.

Or install deliberate, the plugin that ships this one along with the rest of its 11 skills, 1 hook.

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 verify

README.md
[![agentmods](https://agentmods.dev/badge/skills/angad-kandhari/deliberate/verify.svg)](https://agentmods.dev/skills/angad-kandhari/deliberate/verify)
Your own site
<a href="https://agentmods.dev/skills/angad-kandhari/deliberate/verify"><img src="https://agentmods.dev/badge/skills/angad-kandhari/deliberate/verify.svg" alt="Measured on agentmods" height="20"></a>
Per session 69 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,349 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 $0.00069 $0.01349
Opus 5 $0.00034 $0.00674
Sonnet 5 $0.00014 $0.00270
Haiku 4.5 $0.00007 $0.00135

Measured 8d ago against content hash 7e2ff143e55e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, 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 8d 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/verify/SKILL.md · 125 lines

How it starts

The opening of the file, as written. The whole thing — 125 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Verify

Verification skill for LLM coding agents. Load this before claiming a task is done, reporting progress on one, or handing off a change.

Where deliberate keeps code honest and test keeps tests honest, this skill keeps completion claims honest. Counters the most-reported failure mode of agentic coding: work declared done that wasn't, output that looks right but isn't, and a verification burden silently dumped on the human.


1. Done Means Demonstrated

A completion claim is a claim about observed behavior, not written code.

  • Before saying "done," "fixed," or "implemented" — run the thing. The test, the build, the command, the actual flow.
  • Writing code that should work is half the task. Watching it work is the other half.
  • If you can't run it (no environment, missing credentials, no harness), it isn't done — it's written. Say "written, not verified" and name what verification would take.
  • "Should work," "this will now," and "the fix ensures" are predictions. Don't dress predictions as results.

Test: Did I watch this work, or am I predicting it will?


2. Never Game the Check

When the check fails, fix the code — never the check.

The check failing on your change is the check working. Off-limits moves:

  • Deleting or skipping a failing test to get to green.
  • Loosening an assertion until it passes.
  • Hardcoding the expected output.
  • Stubbing the implementation and reporting the feature complete.
  • Catching and swallowing the error that's failing the build.

If you believe the check itself is wrong, stop and make that case out loud. Changing the check is a decision the user makes — not a side effect of reaching green.

Test: After my change, would the check still catch the bug it was written to catch?


3. Green Tests Are Not the Goal

Tests passing is evidence, not the objective. The objective is behavior.

  • After green, exercise the change the way a caller would: run the CLI command, hit the endpoint, load the page, call the function with real-shaped data.
  • Tests verify what tests cover. The gap between "suite passes" and "feature works" is exactly where "almost right" lives.
  • Compiling, typechecking, and linting are necessary, not sufficient. They prove the code parses, not that it works.

Read the full file on GitHub · 125 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. 8d ago First seen · 125 lines · 69 tokens per session scan A 7e2ff143e55e

Subscribe to this mod's changes

verify is a skill published in the GitHub repository angad-kandhari/deliberate (2 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 69 tokens to every session and 1,349 once invoked, about $0.0003 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 2026-08-31.