agentsclimarketplace

Disciplined engineering

Skill xxxxxthhh/disciplined-engineering/skills/disciplined-engineering

Adaptable end-to-end software engineering workflow for planning, implementing, refactoring, debugging, reviewing, and finishing non-trivial code changes with proportional process and evidence-based verification. Use when Codex owns a multi-step engineering outcome, especially when a bug needs systematic diagnosis, work may benefit from delegation, or a change must move safely from requirements to delivery.From its SKILL.md

Install
npx -y skills add xxxxxthhh/disciplined-engineering --skill disciplined-engineering

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.4 KB, 585 tokens by cl100k_base, as published. Nobody here has run it

Disciplined Engineering

Own the outcome from intent through verification. Use the smallest process that provides enough clarity and evidence. Keep one root agent accountable for scope, integration, and delivery.

1. Frame the work

  1. Read repository instructions, current Git state, relevant domain docs, and nearby implementation before proposing work.
  2. Classify the request: explain, diagnose, change, review, or deliver. Do not turn diagnosis or review into an implementation without authorization.
  3. Establish what outcome is required, what evidence would support it, and which delivery actions are allowed.
  4. Handle clear, bounded work directly. Add planning or clarification only when it changes the result.

For ambiguous features, multi-step changes, or large efforts, read planning-and-design.md.

2. Decide proportionally

  • Discover facts from the codebase, tools, and primary sources. Ask the user only for decisions or unavailable facts.
  • Reuse the repository's language, architecture, testing seams, and established workflow.
  • Compare alternatives when a decision is consequential or hard to reverse.
  • Keep scope changes explicit when new evidence changes the work.

3. Delegate only when useful

The root agent owns dispatch, integration, and final verification. Delegate only when the work:

  • is independent and does not overlap shared mutable state;
  • has a bounded deliverable and an objective way to check it;
  • benefits materially from parallel execution or an independent perspective;
  • can be reconciled by the root agent.

Give each child a concrete goal, allowed scope, shared-state constraints, expected artifact, and verification. Do not delegate delegation decisions, final integration, branch completion, or the whole task. Prevent recursive delegation unless the user explicitly requests a hierarchy. Treat child reports as inputs, not proof.

4. Execute through tight feedback loops

  • Prefer small vertical slices and fast observable feedback for behavior changes.
  • Reproduce the reported symptom before fixing a bug when a trustworthy reproduction is feasible.
  • Run narrow checks during the work and preserve unrelated user changes.

Read testing-and-debugging.md when changing behavior or diagnosing a failure.

5. Review and finish

Review the result on separate axes:

  • Spec: required behavior, omissions, incorrect behavior, and scope creep.
  • Standards: repository rules, maintainability, architecture, tests, and security implications.

Read review-and-delivery.md before responding to review feedback, claiming completion, or changing branch/remote state.

Treat this skill as the owner of the general engineering process. Do not invoke another general-purpose engineering workflow for stages already covered here. Use specialized tool, platform, artifact, or security skills only when they add a capability this workflow does not provide, then return control here.

What ships with it: 4 files

4.5 KB alongside SKILL.md

agents/

Keep looking

Skills are one crate of 326,834. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.