ng-classify

A decision aid that chooses how much checking a proposed change needs, from a quick review to a stronger human-reviewed process.

In plain words
What is it for?
Use it before changes to code, documentation, dependencies, AI behavior, releases, or public claims when the appropriate review level is unclear.
Why use it?
It helps match the level of evidence and caution to the change's consequences, reversibility, and unknowns.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/flyfission/nuclear-grade-context-engineering/ng-classify
Clone the repo
git clone --depth 1 https://github.com/FlyFission/nuclear-grade-context-engineering
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 856 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00856
Opus 5 $0.00000 $0.00428
Sonnet 5 $0.00000 $0.00171
Haiku 4.5 $0.00000 $0.00086

Measured 2d ago against content hash 49892e08e688, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ng-classify scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/ng-classify.md · 61 lines

How it starts

The opening of the file, as written. The whole thing — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.

ng-classify

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.

Picks Quick, Standard, or a stronger human-reviewed mode based on consequence, how easy it is to undo, and how much is unknown. Use when you start a change to code, docs, dependencies, AI power, releases, or public claims and the right level of care is unclear. Do not use for a tiny easy-to-undo edit with obvious proof, which is Quick by default, or for a purely administrative reversible edit crossing no trust boundary, which needs no packet at all.

Use when

  • A change request is new, vague, or has grown.
  • You know the decision question, but the evidence bar is unclear.
  • A pull request has AI-generated code, tests, docs, prompts, or release files.
  • Reviewers disagree on whether Quick evidence is enough.
  • The work is routine, by-the-book, new, interrupted, resumed, handed off, or high stakes, and you need the right habit to control it.

Do not use when

  • A change record already has a fresh mode choice and the scope has not changed.
  • The system is failing right now and needs incident handling first.
  • The change is purely administrative, instantly reversible, and crosses no trust boundary -- that is the administrative floor (no packet; the commit message is the record), not a mode to rate.
  • The mode is already chosen and you now need to author or update the packet files -- use creating-change-records instead.

Inputs

  • The user request, issue, pull request, or diff.
  • The files, dependencies, prompts, data, credentials, APIs, release files, and users the change affects.
  • docs/02-operating-system/activation-thresholds.md.
  • The change record's risk.md, if one exists.

Prompt text

Sort this change into a Nuclear-grade mode.

Inputs:
- Request or diff: <paste/link>
- Affected files/assets: <list>
- Impact on users, security, dependencies, data, AI behavior, or release: <known facts>
- Archetype (posture) this work is in, if known: Prototyper / Builder / Sweeper / Grower / Maintainer

Return:
- the decision question and the proof that must clear before work goes on
- the archetype floor that applies (Sweeper and Builder start at Standard; Maintainer at Nuclear on trust-bearing surfaces; a Prototyper's low floor holds only while shipping is an explicit non-goal)
- the chosen mode: Quick, Standard, or a stronger mode that a human reviews -- it must clear that floor, or state the reason for going under it
- how bad it is if wrong, how easy to undo, how exposed, how easy to catch, how uncertain
- the work mode and which safety habit (HPI) to use: none, context pack, handoff, self-check, an independent check, a record of lessons from real operation (OPEX), or a trust check
- the assumptions or facts that drove the mode choice
- the record files this mode needs
- the least proof required
- the conditions that should make you ask for help
- a limits note: do not claim formal verification and validation, compliance, certification, safety, security, or regulatory adequacy

Read the full file on GitHub · 61 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 61 lines · 0 tokens per session scan A 49892e08e688

Subscribe to this mod's changes

ng-classify is a command published in the GitHub repository FlyFission/nuclear-grade-context-engineering (33 stars, last pushed 23d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 856 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.