engineering-discipline

Always-on engineering rules for software work, including tracing the real execution path, defining a runnable check, and keeping one source of truth for data and configuration.

In plain words
What is it for?
Investigating bugs, planning coding tasks, verifying changes, and avoiding duplicated configuration or inconsistent values.
Why use it?
They reduce fixes made in the wrong part of a system and require changes to be checked before they are called complete.

Cursor rule for Cursor

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 rules/mission69b/t2000/engineering-discipline
Clone the repo
git clone --depth 1 https://github.com/mission69b/t2000

Made for: Cursor.

Per session 656 This file is loaded in full into every session.
When invoked 656 The same file — it is already loaded in full.
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 $0.00656 $0.00656
Opus 5 $0.00328 $0.00328
Sonnet 5 $0.00131 $0.00131
Haiku 4.5 $0.00066 $0.00066

Measured 2d ago against content hash 19203021f638, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

engineering-discipline 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.

.cursor/rules/engineering-discipline.mdc · 52 lines

What it actually says

Engineering Discipline

Canonical depth: .claude/skills/t2000-engineering/SKILL.md. That file holds the worked examples, the ask-vs-proceed test, the ESLint flat-config override trap, and the test-file convention. This block is the always-on summary and is intentionally mirrored in CLAUDE.md § Engineering Discipline — those two copies must stay in sync; everything else lives in the skill only.

Replaced (2026-07-24) the four separate always-apply rules engineering-principles · goal-driven-execution · coding-discipline · product-build-algorithm.

Trace before you fix. Trace the ACTUAL execution path (user action → route → handler → SDK → chain → response → UI) and confirm which code actually runs before changing anything. Most multi-iteration fixes are one-iteration fixes that started in the wrong layer.

Verifiable goals, always. Convert every task into a goal with a runnable check. State multi-step plans as step → verify: pairs. Never say "done" without running the verify step; never "should be fine" without re-reading the diff.

Single source of truth. If data exists somewhere, import it. Never copy a token map, decimal, coin type, or config into a second file. Before hardcoding any list, ask whether someone will have to hand-update it later — if yes, the approach is wrong.

Fix at the root. If a fix needs 3+ places or several attempts, the architecture is wrong. Find the single point of failure.

Simplicity first, surgically applied. Minimum code that solves the problem; nothing speculative. No abstractions for single-use code — and none whose shape is shared but whose logic isn't. Touch only what the request requires; match existing style; mention unrelated dead code rather than deleting it.

Remove completely — no orphans. When you delete a feature, sweep every layer in the same pass: source, types, error codes, tests, deps, patches, docs, rules, CI, export barrels, empty dirs. Dead remnants of your own removal are in scope; "keep it just in case" is not.

Run the product algorithm in order. (1) Make the requirements less dumb — requirements are guilty until proven innocent. (2) Delete the part or step; if you aren't adding back ≥10%, you didn't delete enough. (3) Optimize what survived. (4) Accelerate. (5) Automate last. Default to deletion; name the fork rather than silently picking a constrained path.

Machines AND humans. Every surface must work for an autonomous agent and a person. If a design serves only one, it's half-built.

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 · 52 lines · 656 tokens per session scan A 19203021f638

Subscribe to this mod's changes

engineering-discipline is a cursor rule published in the GitHub repository mission69b/t2000 (23 stars, last pushed 2d ago), licensed MIT. It adds 656 tokens to every session, about $0.0033 per session on Opus 5. 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.