nuclear-grade

Portable command prompt generated from `skills/recording-a-known-good-version/SKILL.md`. Edit the skill, then run `python tools/ng.py gen-commands`; do not edit this file by hand.

Portable command prompt generated from `skills/breaking-down-the-work/SKILL.md`. Edit the skill, then run `python tools/ng.py gen-commands`; do not edit this file by hand.

Portable command prompt generated from `skills/rating-change-risk/SKILL.md`. Edit the skill, then run `python tools/ng.py gen-commands`; do not edit this file by hand.

Portable command prompt generated from `skills/closing-stale-packets/SKILL.md`. Edit the skill, then run `python tools/ng.py gen-commands`; do not edit this file by hand.

Portable command prompt generated from `skills/reviewing-code-quality/SKILL.md`. Edit the skill, then run `python tools/ng.py gen-commands`; do not edit this file by hand.

Five subagents map onto the **PROVE** beats, each with tool boundaries that *encode* the authority

PROVE Educate stage. Use after the verdict to lock in the approved baseline and turn operation into a lesson — record the baseline, OPEX/lessons, and any charter update into .nuclear/. Do not use to build product code, decide ship/block, or run the change.

PROVE Verdict stage. Use to make the ship / block / defer / ship-with-named-risk decision on the presented evidence — read-only and context-separated from the runner, but still coupled through the orchestrator and possibly model/resources. Do not use to build, gather new evidence, or write code.

PROVE Observe stage. Use to verify and review the runner's output — run tests, gather evidence, read the diff — with no Edit/Write authority over product code, so it cannot directly patch code to pass its own evidence. Do not use to build, plan, or decide ship/block.

PROVE Plan stage. Use to turn a request into an approved plan — question, discover, specify, plan — writing only to the change packet, never product code. Dispatch first, before any building. Do not use to edit code, run commands, or decide ship/block.

Splits scope into a product-first work breakdown that follows the 100% rule, keeps pieces from overlapping, uses outline numbers, and gives every piece a dictionary entry. Use when an epic, feature, or new subsystem needs a clean split into deliverables, or one source of truth before folders or work begin. Do not use for a one-file edit or a backlog item already broken down.

Prepares focused context for an AI agent, reviewer, verifier, or releaser at the start of a task, with a clear role, goal anchor, authority, evidence to produce, forbidden actions, and stop conditions. Use when an agent or reviewer is about to begin work and needs bounded context before it starts. Do not use for a tiny self-contained task that needs no briefing, or for transferring already-open work to a new owner, which is handing-off-work.

Reviews public text for license, warranty, compliance, safety, security, certification, and fitness claims that go too far, then rewrites them to stay inside the real limits. Use when shipping or editing public docs, READMEs, or rollout copy. Do not use for internal code comments, or for deciding actual legal fitness, which needs a qualified lawyer.

Records a ship, block, defer, or ship-with-risk decision that ties baseline, evidence status, residual risk, rollback, monitoring, and handoff together. Use when a packet, PR, release, dependency change, or agent-authority change approaches merge. Do not use early in development before evidence exists.

Checks that the way you cite source families, agencies, standards, or borrowed ideas is honest and does not claim too much. Use when public docs, templates, skills, change records, or rollout copy point to outside sources. Do not use for private notes, or for checking whether code actually works.