sdd-plan-refactoring
Skill3 starsChanged 35 days agoMITView on GitHub
SDD refactoring when a structural change must preserve behavior; does not implement or add functionality.
In this repository26 other files it gives its agents.
- AGENTS.mdAGENTS.md
- ai-ready-fixSkill
- ai-ready-scoreSkill
- architectural-analysisSkill
- commitSkill
- deep-reviewSkill
- domain-modelingSkill
- dotnet-efficient-validationSkill
- herdr-orchestrationSkill
- no-workaroundsSkill
- repository-cli-efficiencySkill
- sdd-create-prdSkill
- sdd-create-techspecSkill
- sdd-execute-correctionsSkill
- sdd-execute-taskSkill
- sdd-orchestrate-flowSkill
- sdd-orchestrate-prdsSkill
- sdd-orchestrate-tasksSkill
- sdd-plan-auditSkill
- sdd-plan-correctionsSkill
- sdd-plan-tasksSkill
- sdd-review-codeSkill
- sdd-snapshotSkill
- sdd-triageSkill
- update-nuget-packagesSkill
- writing-agents-mdSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Output: every in-scope behavior has an origin and verification, or a concrete question that changes acceptance.
- Very specific: line 16
Output: every R-NN has a proportional safety net and safe sequence.
- Very specific: line 17
When drafting, read assets/TEMPLATE_PRD_REFACTOR.md and assets/TEMPLATE_TECHSPEC_REFACTOR.md in full.
- Distinctive: line 23
Read sources once per version and code only in scope.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Plan SDD refactoring
- Fix target, limits, and slug. A new slug is
[NN]-[name], numbered after the highest folder undertasks/. Resolvetasks/prd-[slug]/prd.mdandtechspec.md. Whencontext-snapshot.mdexists there, load it through the Load branch of.agents/skills/sdd-snapshot/SKILL.mdfirst. Preserve both documents when they exist without update authorization; with it, read current versions and keep IDs. Output: unambiguous operation without implicit overwrite. - Trace inputs, outputs, errors, effects, callers, persistence, integrations, and edges of the target. Use code, tests, logs, fixtures, and contracts as evidence; other features' SDD artifacts are neither evidence nor updated. When the target has many callers or integrations, send read-only explorers in parallel with one question each and verify the lines they cite. Number behaviors
R-01,R-02; distinguish intent, apparent defect, and gap. Output: every in-scope behavior has an origin and verification, or a concrete question that changes acceptance. - Plan characterization only for behavior without protection; prefer semantic assertions. A snapshot test requires a reviewed baseline. Sequence characterization before mutation and reversible slices before irreversible changes.
When the target includes provider adapters, credential or database readers, rate-limit handling, or HUD windows, read references/providers-and-hud.md in full. For WPF views, view models, or bindings, read references/wpf.md in full. In desktop C#/.NET, omit E2E; record existing unit and integration projects, runner, and commands, plus a manual script for uncovered visual behavior. Record an unavailable environment as pending.
Output: every
R-NNhas a proportional safety net and safe sequence; new behavior is separate from scope. - When drafting, read assets/TEMPLATE_PRD_REFACTOR.md and assets/TEMPLATE_TECHSPEC_REFACTOR.md in full. Write only the two artifacts: behavior and acceptance in the PRD; decisions, components, tests, and rollback in the TechSpec. Reference IDs without copying requirements.
Fill in the quality profile with today's measure and the target at the end — in a refactoring the baseline is the target to reduce, and without a number there is no way to prove the structure improved. When this refactoring prepares the terrain for a feature, the target is what makes that change easy, not perfection of the target code.
Output: documents cover every
R-NN, have no placeholders, include a measured profile and a declared target, and servesdd-plan-tasks. - Report paths, risks, gaps, and impacts on derivatives. In standalone use, point to task planning and run the session pause from that reference with
sdd-plan-tasksas next step; in the orchestrated flow, return to behavior HIL before decomposition. Output: reviewable contracts, with no code change or silent promotion of decisions.
Read sources once per version and code only in scope. A behavior contradiction or missing critical verification blocks affected items until a human decision; preserve evidence.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

