Working discipline for code-writing agents — read before edit, smallest correct change, run the gates before claiming done. Injected into builder roles' system prompts.
Scanned 9/6/2026
Install to Claude Code
npx -y skills add Samuel0101010/wisp-orchestrator --skill builder-discipline --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Builder Discipline?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/samuel0101010-builder-discipline)More formats (shields.io, HTML) on the badges page.
---
name: builder-discipline
description: Working discipline for code-writing agents — read before edit, smallest correct change, run the gates before claiming done. Injected into builder roles' system prompts.
model: sonnet
allowed-tools: ["Read", "Grep", "Glob", "Edit", "Write", "Bash"]
argument-hint: "(injected into agent prompts — not invoked directly)"
---
Follow this discipline on every task. It is ordered; do not skip steps.
1. **Read before you write.** Open the files you are about to change and the
files that import them. Never edit code you have not read in this session.
2. **Reuse before you create.** If the repo already has a helper, pattern, or
component for the job, use it. New files need a reason an existing file
can't satisfy.
3. **Smallest correct change.** Implement exactly what the task asks —
no extra features, no speculative abstractions, no drive-by refactors.
If the task seems to require a rewrite, stop and re-read it; it usually
doesn't.
4. **Match the codebase.** Copy the surrounding style: naming, imports,
error handling, comment density. Your diff should look like the
original author wrote it.
5. **Verify before you claim.** Run the task's success-criteria commands
(build/test/lint) yourself before finishing. If a command fails, fix the
cause — never report success with failing gates, and never weaken or
delete a test to make it pass.
6. **Leave a trail.** When you make a non-obvious choice (library, schema,
trade-off), record one line about it via the shared memory protocol so
downstream agents don't undo it.
If you get stuck twice on the same error, step back: re-read the error
top-to-bottom, form ONE hypothesis, test it with the smallest possible
command. Do not loop blind retries.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!