Verification Before Completion
verification-before-completion/
Stops Claude from claiming work is done, fixed or passing until it has run the check and read the output. Evidence before assertions.
Evidence before assertions, run the check and read its output before claiming anything works.
Why use it
- "Should work", "this fixes it", "tests pass", stated without ever running the command.
- A confident wrong claim costs more than a slow right one; it gets committed, merged, and shipped.
What it does
- Blocks any completion / fixed / passing claim until the verification command has actually run.
- Requires the real output, not a summary of what the output would presumably be.
- Forces an explicit statement of what remains UNVERIFIED when a check cannot be run.
Benefits
- Trust the green. A pass means a command ran and you can read its output.
- Catch it before the commit: not after the deploy, at 2am.
- Kill the confident wrong answer, the most expensive failure an agent produces.
- One file. Any stack. Works with whatever test runner you already have.
Install
Download this skill and copy the verification-before-completion folder to .claude/skills/verification-before-completion/ in your project, or to ~/.claude/skills/verification-before-completion/ to use it in every project. Claude Code finds it on its own; you can also call it with /verification-before-completion.
Part of the Agent Engineering bundle. To get the whole set as a Claude Code plugin: /plugin marketplace add endorphin-ai/hasbrains-agent-kit, then /plugin install agent-engineering@hasbrains-agent-kit.
Source: endorphin-ai/hasbrains-agent-kit. License: MIT.
This skill ships no assets.