verify-this

Verifies claims by comparing baseline and treatment evidence into a verdict.

Updated Nov 10, 2013
One-click install
npx skills add https://github.com/bnferguson/dotfiles --skill verify-this-bnferguson
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: verify-this
Source: https://github.com/bnferguson/dotfiles/tree/main/.agents/skills/verify-this
Command: npx skills add https://github.com/bnferguson/dotfiles --skill verify-this-bnferguson

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It turns vague claims like "this fix works" into falsifiable statements backed by repeatable local evidence, so you get a clear verdict instead of a recap. ## Core Features & Use Cases - Falsifiable Claim Restatement: Converts a claim into a condition, metric, and threshold before any testing begins. - Baseline vs Treatment Comparison: Captures evidence from the old state and the changed state using the same command, data, and environment. - Structured Verdict Output: Returns exactly one of VERIFIED, NOT VERIFIED, or INCONCLUSIVE with metrics, deltas, and reasoning. - Use Case: After fixing a performance regression, ask for verification and get a side-by-side timing comparison of the parent commit versus the fix branch with a definitive verdict. ## Quick Start Ask the agent to verify that your recent fix actually resolves the bug by comparing behavior before and after the change.

Frequently Asked Questions about verify-this

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

FAQPage Schema
How do I verify that a bug fix actually works?▼

Restate the fix as a falsifiable claim with a metric and threshold, then capture baseline behavior from the parent commit or broken repro and treatment behavior from the fixed state. Compare the artifacts and return VERIFIED only if the change moves in the predicted direction by the claimed threshold.

How to measure performance before and after a code change?▼

Run the same timing or profiling command on the baseline state and the changed state with identical data, warmup, and environment. Compare the numbers against the claimed threshold; unchanged or noisy results yield NOT VERIFIED or INCONCLUSIVE.

What evidence counts when verifying a UI or CLI behavior claim?▼

Acceptable evidence includes terminal transcripts, screenshots, accessibility snapshots, browser traces, HTTP response diffs, heap snapshots, and test output. The key requirement is that baseline and treatment artifacts come from the same command and environment.

When is a verification result inconclusive?▼

A result is INCONCLUSIVE when no valid baseline exists, the signal is too noisy, the measurement fails, or an environment difference invalidates the comparison. It does not mean the claim is false, only that the evidence cannot decide it.

Can verification artifacts be saved to disk?▼

Yes, artifacts can be stored under /tmp/verify-this/<claim-slug>/ with claim, timeline, baseline, treatment, diff, and verdict files. If artifacts contain sensitive code, screenshots, or heap data, only minimal inline evidence is kept unless the user agrees to disk storage.