verification-planning

Plans evidence paths for verifying non-trivial code changes before implementation.

1|Updated Jul 23, 2026
One-click install
npx skills add https://github.com/sanjanb/my-agent-harness --skill verification-planning-sanjanb
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: verification-planning
Source: https://github.com/sanjanb/my-agent-harness/tree/main/skills/verification-planning
Command: npx skills add https://github.com/sanjanb/my-agent-harness --skill verification-planning-sanjanb

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Non-trivial code changes often ship without a credible way to prove they work, leaving teams guessing whether a feature, bug fix, or refactor actually behaves as claimed. This Skill structures the upfront planning of how a change will be verified, so every claim has a concrete evidence path before implementation begins. ## Core Features & Use Cases - Evidence Path Design: Frames the claim being made, then derives project-specific ways to establish, limit, or refute it using the system's inputs, outputs, state transitions, and boundaries. - Verification Affordances: Guides creation of minimal capabilities that make relevant state controllable, observable, repeatable, and diagnosable when existing surfaces are insufficient. - Verification Budgeting: Assigns one owner per distinct claim and selects the minimum non-duplicative evidence covering claims and important boundaries. - Use Case: Before refactoring a payment retry flow, use this Skill to define the exact claims (e.g., no duplicate charges), choose how to observe them, and decide what test support is needed before writing any code. ## Quick Start Ask the agent to build a verification plan for the upcoming feature or bug fix, framing the claim and evidence path before any implementation starts.

Frequently Asked Questions about verification-planning

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

FAQPage Schema
How do I plan verification before implementing a code change?

Start by framing the claim: state the behavior that must become true and the conditions that could make a confident conclusion wrong. Then derive evidence paths from the system's controllable inputs, observable effects, and state transitions, choosing the one that produces a trustworthy conclusion at proportionate cost.

What is an evidence path in software verification?

An evidence path is a project-specific route from a claim about system behavior to evidence that can establish, limit, or refute it. It is derived from the system's own inputs, outputs, invariants, boundaries, and ability to repeat or reverse a scenario rather than from a generic testing technique.

When should I use verification planning versus just running tests?

Use verification planning for non-trivial work: features, bug fixes, refactors, cross-system changes, or behavior changes needing credible proof. Small mechanical changes can follow ordinary project checks directly without a formal evidence path.

What is a verification affordance and when do I need one?

A verification affordance is the smallest capability that makes relevant system state controllable, observable, repeatable, and diagnosable. Create one when the existing system leaves the decisive truth too indirect or ambiguous, and decide deliberately whether it is temporary or durable.

How do I avoid over-testing or duplicating verification effort?

Set a verification budget: state the distinct claims, assign one owner per claim, and choose the minimum non-duplicative evidence covering claims and boundaries. Reuse evidence only while its relevant code, inputs, environment, and state remain valid.