ptp-verification-before-completion

Validates completion claims against fresh, risk-proportional evidence from the current run.

Updated Jun 8, 2026
One-click install
npx skills add https://github.com/AlmogMaayan/ptp --skill ptp-verification-before-completion-almogmaayan
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ptp-verification-before-completion
Source: https://github.com/AlmogMaayan/ptp/tree/main/skills/ptp-verification-before-completion
Command: npx skills add https://github.com/AlmogMaayan/ptp --skill ptp-verification-before-completion-almogmaayan

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It prevents premature or unsupported claims that work is complete, fixed, or passing by requiring every completion claim to be backed by fresh evidence produced in the current run. ## Core Features & Use Cases - Evidence-gated completion claims: No success or "fixed" claim is admitted without a command and its result from the current run. - Risk-proportional verification: The depth of checking scales with the blast radius of the claim being made. - Explicit gap and N/A reporting: Unresolved gaps withhold the completion claim, while genuinely inapplicable checks are reported as N/A: <reason> rather than passes. - Use Case: After fixing a bug and re-running the test suite, report the claim, the exact test command, and its passing output instead of saying the fix "should work". ## Quick Start Before claiming any task is done, apply the verification-before-completion rules and report each claim with the command that proves it and that command's result.

Frequently Asked Questions about ptp-verification-before-completion

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

FAQPage Schema
How do I verify work is complete before claiming it is done?

Run a concrete check in the current session, such as a test suite or build command, and report the claim alongside the command and its result. Claims based on assumption or wording like "should work" are not accepted as evidence.

What counts as evidence for a completion claim?

Evidence is the output of a command executed in the current run that directly establishes the claim, proportional to the claim's risk. Evidence produced before the most recent edit is stale and must be re-run.

Can I rely on a delegated agent's success report?

No. A delegated run's own success report is not evidence. You must inspect the resulting files or diff yourself before making any completion claim about the delegated work.

What should I report when a verification check does not apply?

Report it as `N/A: <reason>`, which is distinct from both a pass and an unresolved gap. Never use N/A where a runnable check exists, and never treat it as evidence of success.

What happens when verification cannot be fully completed?

State the unresolved gap explicitly and withhold the completion claim, reporting only the partial verified state. Hedging language like "probably" or "seems to" is not a substitute for the missing check.