verify

Enforces fresh execution of verification commands before any completion claim.

1|Updated May 1, 2026
One-click install
npx skills add https://github.com/FerEscobarDev/Specture --skill verify-ferescobardev
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: verify
Source: https://github.com/FerEscobarDev/Specture/tree/main/skills/verify
Command: npx skills add https://github.com/FerEscobarDev/Specture --skill verify-ferescobardev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI agents and developers often claim work is done, tests pass, or bugs are fixed without actually running the verification command in the current turn, producing broken code and eroding trust. This Skill acts as a mandatory gate that blocks any completion claim until fresh evidence exists. ## Core Features & Use Cases - Iron Law Gate: Blocks claims like "done", "fixed", or "passing" unless the verification command was run in the current turn and its output was read. - Claim-to-Verification Map: Maps common claims (tests pass, build succeeds, bug fixed, epic complete) to the exact command and output required to prove them. - Rationalization Detection: Lists red flags and common excuses ("should work", "previous run passed") that signal an unverified claim is about to be made. - Use Case: Before marking an epic complete in a ROADMAP file, run the full test suite, linter, and type-check per the project stack configuration, then only write the completion marker if all checks pass. ## Quick Start Before marking this task complete, run the test suite and linter fresh and show me the output as evidence.

Frequently Asked Questions about verify

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify code changes before claiming a task is done?▼

Identify the command that proves the claim, run it fresh in the current turn, read the full output and exit code, then confirm the output actually supports the claim. Only state completion after this sequence, citing the evidence.

What counts as sufficient verification evidence for tests passing?▼

Sufficient evidence is running the test command now and seeing zero failures in the current output. Previous runs, confidence, partial checks, or another agent's success report do not qualify as evidence.

Does a passing linter prove the build succeeds?▼

No, a linter is not the compiler. Linting only checks style and static issues, so you must run the actual build command and confirm exit code 0 before claiming the build succeeds.

When should I not use a verification gate before completion?▼

If the verification command fails due to environment issues unrelated to the work, route to debugging instead. If verification is impossible, such as hardware-specific features, state explicitly what can and cannot be verified rather than claiming completion.

Why do AI agents claim work is done without running tests?▼

Agents often rationalize with excuses like "should work now" or trust prior runs and other agents' reports. The verification gate counters this by requiring fresh command execution and output reading before any success claim.