From stn-skills
Use for end-to-end feature delivery from idea to verified code. Orchestrates brainstorming, plan-writing, and plan-execution with validated handoffs and 11 gates. No shortcuts, no skipped gates, no unverified claims. Triggers: "build feature", "build this", "implement end-to-end", "full pipeline".
How this skill is triggered — by the user, by Claude, or both
Slash command
/stn-skills:build-featureThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Meta-orchestrator that chains three complete workflows — brainstorming, plan-writing, and plan-execution — into a single design-to-delivery pipeline. Each macro-phase runs its full sub-workflow with all phases and gates preserved. Pipeline state transfers between macro-phases via file paths on disk, making the pipeline resumable across sessions.
Meta-orchestrator that chains three complete workflows — brainstorming, plan-writing, and plan-execution — into a single design-to-delivery pipeline. Each macro-phase runs its full sub-workflow with all phases and gates preserved. Pipeline state transfers between macro-phases via file paths on disk, making the pipeline resumable across sessions.
Core principle: From idea to verified code in one pipeline. No shortcuts. No skipped gates. No unverified claims. No deprecated code. No legacy patterns.
Announce: "I'm using the build-feature skill to run the full design-to-delivery pipeline."
FROM IDEA TO VERIFIED CODE IN ONE PIPELINE.
NO SHORTCUTS. NO SKIPPED GATES. NO UNVERIFIED CLAIMS.
graph TD
M1["Macro-Phase 1: Design\n(brainstorming)\n6 phases, 4 gates"] --> T1["docs/specs/YYYY-MM-DD-topic-design.md"]
T1 --> V1{"Handoff\nValidator\n(Mode A)"}
V1 -->|validated| M2["Macro-Phase 2: Plan\n(plan-writing)\n6 phases, 4 gates"]
M2 --> T2[".plan/plan-YYYYMMDD-slug.md"]
T2 --> V2{"Handoff\nValidator\n(Mode B)"}
V2 -->|validated| M3["Macro-Phase 3: Execute\n(plan-execution)\n7 phases, 3 gates"]
M3 --> Done(("Completion Report\n+ Fidelity Score"))
T1 -.->|"user may exit"| E1(("Design complete"))
T2 -.->|"user may exit"| E2(("Plan complete"))
classDef macro fill:#2563eb,stroke:#1d4ed8,color:#fff,font-weight:bold
classDef artifact fill:#16a34a,stroke:#15803d,color:#fff
classDef validator fill:#d97706,stroke:#b45309,color:#fff,font-weight:bold
classDef done fill:#7c3aed,stroke:#6d28d9,color:#fff
classDef exit fill:#6b7280,stroke:#4b5563,color:#fff
class M1,M2,M3 macro
class T1,T2 artifact
class V1,V2 validator
class Done done
class E1,E2 exit
State transfers between macro-phases via files — no in-memory coupling:
| Transition | Artifact | Location |
|---|---|---|
| Design to Plan | Design spec | docs/specs/YYYY-MM-DD-<topic>-design.md |
| Plan to Execute | Implementation plan | .plan/plan-{YYYYMMDD}-{slug}.md |
This makes every macro-phase independently resumable. If a session ends after Design, start a new session at Plan by pointing to the spec file.
At the start of every session, read .claude/stn-skills-pipeline-state.json if it exists.
active_skill via the Skill tool. The skill's Session Resumption Protocol reads the state file and continues from the correct phase.pipeline_id: "{YYYYMMDD}-{slug}", active_skill: "brainstorming", current_phase: 1, total_phases: 6, gates_passed: [], gates_total: 4, artifact_path: null, handoff_validated: false. Begin from Macro-Phase 1.This ensures build-feature survives session boundaries without losing progress or restarting completed work.
How to execute: Read skills/brainstorming/SKILL.md and follow its complete 6-phase workflow from Phase 1 through Phase 6, including all 4 gates. Dispatch agents from skills/brainstorming/agents/ and load references from skills/brainstorming/references/ exactly as that SKILL.md specifies.
skills/brainstorming/agents/ (problem-decomposer, assumptions-surfacer, multi-lens-explorer, approach-evaluator, adversarial-reviewer)skills/brainstorming/references/ (cognitive-lenses, decision-matrix-template, design-spec-template, reasoning-flaw-catalog)docs/specs/YYYY-MM-DD-<topic>-design.mdError handling: If adversarial review finds Blockers (Phase 4), resolve them within brainstorming — do NOT escalate to build-feature level. The brainstorming SKILL.md defines its own blocker resolution loop.
MANDATORY: After GATE 4, invoke the next skill via the Skill tool. Do NOT start planning without it.
After GATE 4 (Final Spec Approval), use AskUserQuestion:
{path}. Continue to plan-writing, or stop here?"On "Continue to plan-writing": Immediately invoke the Skill tool: Skill(skill: "stn-skills:plan-writing", args: "{spec_file_path}")
On "Stop here": The pipeline ends. The design spec is on disk and can be used independently with /stn-skills:plan-writing in a future session.
Before starting Macro-Phase 2, run skills/pipeline-handoff-validator/SKILL.md Mode A on the design spec file. Present the Handoff Compliance Table. If gaps are found, offer to return to brainstorming or proceed with acknowledged gaps.
How to execute: Read the design spec file from Macro-Phase 1. Then read skills/plan-writing/SKILL.md and follow its complete 6-phase workflow. In Phase 1, pass the design spec file path as input. Dispatch agents from skills/plan-writing/agents/ and load references from skills/plan-writing/references/ exactly as that SKILL.md specifies.
skills/plan-writing/agents/ (codebase-cartographer, task-decomposer, step-author, plan-verifier)skills/plan-writing/references/ (plan-document-template, task-anatomy, placeholder-detector-rules).plan/plan-{YYYYMMDD}-{slug}.mdError handling: If Plan Quality Score < 90 after 2 rework cycles (Phase 5), present remaining defects to user at GATE 3. User decides: accept with known gaps, or stop pipeline.
MANDATORY: After GATE 4, invoke the next skill via the Skill tool. Do NOT start executing without it.
After GATE 4 (Final Plan Approval), use AskUserQuestion:
{path}. Continue to execution, or stop here?"On "Continue to execution": Immediately invoke the Skill tool: Skill(skill: "stn-skills:plan-execution", args: "{plan_file_path}")
On "Stop here": The pipeline ends. Both spec and plan are on disk. Resume execution later with /stn-skills:plan-execution pointing to the plan file.
Before starting Macro-Phase 3, run skills/pipeline-handoff-validator/SKILL.md Mode B on the plan file. Present the Handoff Compliance Table. If gaps are found, offer to return to plan-writing or proceed with acknowledged gaps.
How to execute: Read the plan file from Macro-Phase 2. Then read skills/plan-execution/SKILL.md and follow its complete 7-phase workflow. In Phase 1, pass the plan file path as input. Dispatch agents from skills/plan-execution/agents/ and load references from skills/plan-execution/references/ exactly as that SKILL.md specifies.
skills/plan-execution/agents/ (task-implementer, spec-compliance-reviewer, code-quality-reviewer, integration-reviewer, completion-verifier)skills/plan-execution/references/ (checkpoint-protocol, circuit-breaker-thresholds, completion-report-template, drift-detection-rules, reflect-retry-escalate, status-codes, task-handoff-template)Error handling: Circuit breakers (YELLOW/RED) and adaptive replanning are handled within plan-execution's Phase 3. If execution halts (RED circuit breaker), the .claude/plan-execution-state.json state file preserves progress for resumption.
/stn-skills:plan-writing with spec path. After Plan: /stn-skills:plan-execution with plan path.If you catch yourself:
ALL of these mean: STOP. The pipeline exists because shortcuts produce failures.
| Excuse | Reality |
|---|---|
| "The user already has a clear idea, skip brainstorming" | Clear ideas still have unexplored alternatives and hidden assumptions. Use Focused complexity. |
| "The spec is simple, skip handoff validation" | Simple specs with missing acceptance criteria produce the most rework during execution. |
| "I can plan and execute simultaneously" | Simultaneous planning and execution eliminates the verification gate that catches plan defects. |
| "Handoff validation is redundant after GATE 4 approval" | User approval validates direction. Contract validation checks completeness. Different concerns. |
| "This is a small feature, the full pipeline is overkill" | Small features through the full pipeline take 30 minutes. Small features debugged after skipping gates take hours. |
| "I'll remember the spec details, no need to read the file" | Memory drifts. The file is the contract. Read it. |
npx claudepluginhub sthiermann/stn-skills --plugin stn-skillsPlans single features, endpoints, or infrastructure via Socratic conversation to uncover requirements, edge cases, failure modes, and constraints; outputs markdown feature plan artifact before coding.