AGENT SKILL

verify-and-explain

Audit a claim or implementation, prove what can be proven with concrete evidence, and explain the result in clear audience-appropriate language. Use for reviews, validation summaries, change explanations, test reports, and technical findings that must distinguish fact from inference.

License 许可
MIT
最近更新
2026-11-13
Stars
★ 4
Forks
3

01 前言 Front Matter

name
verify-and-explain
engines
[ "AGNOSTIC" ]
description
Audit a claim or implementation, prove what can be proven with concrete evidence, and explain the result in clear audience-appropriate language. Use for reviews, validation summaries, change explanations, test reports, and technical findings that must distinguish fact from inference.

02 资源文件 2

03 SKILL.md 正文

Verify And Explain

When to use

  • A user asks whether work is correct, complete, safe, or functioning.
  • A technical result must be explained to a non-specialist.
  • Review evidence is spread across tests, logs, code, screenshots, or configuration.

Workflow

  1. Restate the claim and define what evidence would prove or disprove it.
  2. Inspect authoritative code, tests, logs, artifacts, and runtime behavior.
  3. Label each conclusion as verified, contradicted, inferred, or unknown.
  4. Resolve contradictions before summarizing; do not average conflicting evidence.
  5. Explain the outcome from impact to cause to proof to next action.
  6. Match vocabulary and depth to the named audience without removing material risk.
  7. For multi-condition claims, use the acceptance and evidence matrix to make coverage, freshness, contradictions, and unknowns explicit.
  8. For checkpointed ordinary tasks, follow the execution contract at skills/create-task-file/references/execution-contract.md. Run commands/validate-task-checkpoint.sh <task> --phase resume before trusting a handoff and --phase complete before completion/readiness. Inspect external observations again when their freshness matters.

Output

  • A direct result and confidence level.
  • A compact evidence list with commands or artifacts.
  • Unknowns, residual risk, and the smallest useful next action.

Guardrails

  • Do not present successful command execution as proof of unrelated behavior.
  • Do not invent measurements, test results, or causal explanations.
  • Use plain language without becoming vague or patronizing.
  • Checkpoint validation proves only the declared file identities and mechanical conditions. It does not judge semantic correctness, independently verify an external system, or grant publication approval.
  • Changed candidates invalidate affected proof, including dirty and untracked outputs. Re-run affected checks; merely re-recording hashes must not turn stale evidence into verified evidence.
  • Preserve failed attempts and explain the evidence or changed conditions that justify a retry. Never lower acceptance criteria to make a check pass.

04 同仓库更多 Skills

TimothyMeadows
TimothyMeadows/Hephaera

verify-web3-game-lifecycle

Verify a game's selected Web3 asset, wallet, ownership and trading lifecycle with contract properties, canonical recovery, actual target evidence and separate release readiness.

★ 4
TimothyMeadows
TimothyMeadows/Hephaera

verify-onchain-game-assets

Verify exact on-chain game collection ownership and quantities at canonical blocks, reconcile inventory events, and design honest match admission. Use for ERC721/ERC1155 inventory, reorg recovery, reservations, or transfer-during-match rules.

★ 4
TimothyMeadows
TimothyMeadows/Hephaera

plan-web3-game-assets

Define an opt-in Ethereum, Base or Immutable game-asset lifecycle, including token identity, supply, rights, custody, game eligibility and chain-specific dependencies.

★ 4
TimothyMeadows
TimothyMeadows/Hephaera

integrate-game-wallets

Integrate explicit player-wallet connection, scoped SIWE authentication, authority invalidation and reviewed target-specific handoff without application-held player keys.

★ 4
TimothyMeadows
TimothyMeadows/Hephaera

integrate-game-asset-trading

Integrate explicitly selected game asset settlement with exact unsigned intents, scoped approvals, idempotent operations and canonical receipt reconciliation. Use for NFT listing, buying, cancellation, retries or marketplace integration.

★ 4
TimothyMeadows
TimothyMeadows/Hephaera

build-evm-game-contracts

Author and verify host-owned EVM game collections and explicit mint, burn, craft or admin rules with pinned Solidity tooling. Use for contract source and deployment candidates; deployment and player signing require separate authorization.

★ 4