evidence-packet

name	evidence-packet
description	Produce concise, reviewable evidence packets for completed, blocked, or owner-gated lifecycle work.

Evidence Packet

Use When

Use this skill whenever an agent completes work, blocks on a missing input, requests owner approval, or hands off to another runtime lane.

Inputs

• target app
• lifecycle stage
• work packet
• actions taken
• artifacts changed
• verification output
• blockers or approval gates

Outputs

• target app and stage
• objective
• inputs used
• changed artifacts
• checks run and result
• claim limits
• blockers
• owner approval request if needed
• next action

Rules

• Evidence should be reproducible from files and commands, not private chat memory.
• Keep the packet short enough to review quickly.
• Link or name artifacts instead of pasting long logs.
• State omissions plainly.
• Do not mark owner-gated work complete until approval evidence exists.

Stop Conditions

• The evidence would require publishing private logs, credentials, or paths.
• The verification result contradicts the intended claim.
• The next action is ambiguous.

Verification

A reviewer should be able to answer: what changed, how it was checked, what is still blocked, and what happens next.