varietyz/disciplined-ai-software-development/reasoning-templates
Claude Code and agent skill by varietyz/disciplined-ai-software-development
- community
Checking that the SKILL.md file is still in the repository.
Execute the PAG reasoning protocols in .{provider}/templates/ when producing a plan, checklist, task breakdown, workflow, agent, or verification flow. Mandatory in plan mode - draft, then compare against the checklist template, then restructure. Use before writing any structured construct, not after.
Install
npx vskill@latest install varietyz/disciplined-ai-software-development/reasoning-templatesSKILL.md
reasoning-templates
The templates in .{provider}/templates/ are executable protocols for reasoning, not reference docs. Producing a structured construct means walking the template's ten-node loop as stated.
The full rule is .{provider}/rules/template.rule.md.
Which template
The dispatch key is each template's LAYER 0 genesis question.
| producing | template | |
|---|---|---|
| plan, checklist, task breakdown | checklist | |
| verdict on claims, audit, context check | claim | |
| shared abstraction from repeated evidence | pattern | |
| reusable specialized investigator | agent | |
| a template from an executed document | layer |
Plan mode: three steps, never one
- Draft the plan from the task.
- Read `checklist` and compare the draft to it, node by node and gate by gate.
- Restructure the plan to what the template enforces.
Step 3 is never skipped on the grounds that the draft already covers it. The draft is raw material, and the template is the authority.
What the restructure adds
An ordinary plan lacks all of these:
- Worth before work: the branch is
argmax(utility − cost)over admissible branches, with the ranking shown,
rather than the first workable idea.
- Topology ordering: dependency-topological then genesis. Severity routes failure and never groups or orders
phases.
- Genesis consistency: no phase depends on a later-genesis output than it produces. Such an inversion is a
decomposition defect, not a tie to break.
- Four dependency axes per phase: sequential, lateral, diagonal and propagation.
- Ripple chains: affected entities carried as names, never counts. An empty dimension carries its
applicability evidence rather than being omitted.
- Evidence contracts: every material claim carries evidence, a named refuter, and confidence at or above
threshold. "No contradiction found" is not support.
- Gate separation: generation-time gates resolve during planning, and future execution gates ship
unchecked.
- Explicit termination: saturation and completion and verification. A self-assessed "done" does not
terminate.
The four gates always run
tel-priority · constrain · ver-stop · ter-stop. The epistemic axes are selectable, and these four are not. A gate naming no evidence is ceremony and fails.
Execution discipline
Follow as stated. Node order, contracts, typed decisions and invariants are not condensed or partially applied.
The mechanism transfers, and catalogs re-derive. The loop, typing and gates are domain-neutral. Where a template's catalog assumes a construct this tree does not carry, such as a class hierarchy, a migration surface, a build step or a registry, that catalog is re-derived against the constructs that are here, and where the slot naming one resolves ABSENT, its branch does not run. Substituting the mechanism is the violation, and re-deriving a catalog is the work. An example catalog is the half a token sweep cannot see, so it is found by reading the protocol whole rather than by searching it.
Executed only. A descriptive artifact (reference, spec, contract, note) is not full-shaped, and forcing the loop onto it is the Procrustean failure.
Never consolidate the templates. Each inlines the full structure so it stays independently executable, which is the deliberate exception to the duplication rule.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingcoordination.surface/.{provider}/skills/reasoning-templates/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/varietyz/disciplined-ai-software-development/reasoning-templates)