Use when you have a written implementation plan to execute in a separate session with review checkpoints
spec
1
0
# Install this skill:
npx skills add miles-knowbl/orchestrator --skill "spec"
Install specific skill from multi-skill repository
# Description
Compile requirements into production-ready technical specifications. Transforms requirements documents into comprehensive 18-section FeatureSpecs (~2000+ lines) with full code, SQL migrations, UI components, feedback timing, and senior engineer audit. Maintains Process Map as system-of-record. The primary driver of the implementation process.
# SKILL.md
name: spec
description: "Compile requirements into production-ready technical specifications. Transforms requirements documents into comprehensive 18-section FeatureSpecs (~2000+ lines) with full code, SQL migrations, UI components, feedback timing, and senior engineer audit. Maintains Process Map as system-of-record. The primary driver of the implementation process."
phase: INIT
category: core
version: "1.0.0"
depends_on: ["architect"]
tags: [planning, specification, requirements]
Spec
Compile requirements into production-ready technical specifications.
When to Use
- After requirements are gathered β Requirements doc is approved, ready for technical spec
- Before implementation β Need production-ready specification with full code
- Complex feature β Multiple entities, services, UI components
- Team handoff β Spec becomes the implementation contract
- When you say: "compile this spec", "write the technical spec", "create a FeatureSpec"
Reference Requirements
MUST read before applying this skill:
| Reference | Why Required |
|---|---|
18-section-template.md |
Defines all required sections and format |
capability-format.md |
Structure for each capability definition |
Read if applicable:
| Reference | When Needed |
|---|---|
feedback-framework.md |
For UI-heavy features requiring timing/haptics |
process-map-template.md |
When maintaining cross-spec tracking |
memory-axioms.md |
For state management patterns |
Verification: Ensure your FeatureSpec has all 18 sections before marking complete.
Required Deliverables
| Deliverable | Location | Condition |
|---|---|---|
FEATURESPEC.md |
Project root | Always (1500+ lines minimum) |
| Process Map update | domain-memory/{domain}/process-map.md |
If Process Map exists |
Core Concept
Spec compilation answers: "How exactly do we build this?"
A compiled FeatureSpec is:
- Complete β 18 sections covering all aspects of implementation
- Production-ready β Full code, not pseudocode or placeholders
- Audited β Senior engineer review with issues resolved
- Traceable β Numbered capabilities, entities, and specs
A FeatureSpec is NOT:
- Requirements (that's requirements)
- High-level architecture (that's architect)
- The actual implementation (that's implement)
The Compilation Loop
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β SPEC COMPILATION LOOP β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β ββββββββββββββββ ββββββββββββββββ ββββββββββββββββ β
β βRequirements β β Compiler β βFeatureSpec β β
β βv1.0 (PM) ββββββΆβ Process ββββββΆβ v1.1 β β
β β~200-500 linesβ β (Claude) β β ~2000+ lines β β
β ββββββββββββββββ ββββββββββββββββ ββββββββββββββββ β
β β β β
β β β β
β βΌ βΌ β
β βββββββββββββββββββββββββββββββββββ β
β β Process Map β β
β β β’ Entities registry β β
β β β’ Coverage tracking β β
β β β’ Compiled specs list β β
β β β’ Architecture diagram β β
β βββββββββββββββββββββββββββββββββββ β
β β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
Compilation Workflow
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β TEMPLATE-BASED COMPILATION β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β INPUT β
β βββββ β
β β’ Requirements document (from `requirements` skill) β
β β’ This compilation process β
β β’ Exemplar specs (previous compilations) β
β β’ Process Map (current version) β
β β
β βΌ β
β β
β STEP 1: Structure Against Template β
β ββββββββββββββββββββββββββββββββββ β
β β’ Create all 18 sections from template β
β β’ Assign Spec ID (SPEC-###) β
β β’ Number capabilities (CAP-001, CAP-002, ...) β
β β’ Fill header with compilation summary table β
β β’ Reference exemplar specs for format/style β
β β
β βΌ β
β β
β STEP 2: Apply Feedback Framework β
β ββββββββββββββββββββββββββββββββ β
β β’ Add Feedback Timing Requirements section β
β β’ Add feedback block to each capability β
β β’ Add haptic patterns, timing bands, animations β
β β’ Implement optimistic UI in all service layer code β
β β’ Apply Paintbrush Test to every interaction β
β β
β βΌ β
β β
β STEP 3: Apply Production Framework β
β βββββββββββββββββββββββββββββββββ β
β β’ Senior Engineer Audit (document issues found) β
β β’ Integrate ALL fixes into spec (not comments) β
β β’ Add observability (metrics, logs, alerts) β
β β’ Add test scenarios (security, concurrency, failure, load) β
β β’ Add feature flags and rollback plan β
β β
β βΌ β
β β
β STEP 4: Validate Template Compliance β
β ββββββββββββββββββββββββββββββββββββ β
β β’ Check all 18 sections present β
β β’ Verify minimum 1,500 lines (target 2,000+) β
β β’ Verify numbered capabilities with feedback β
β β’ Verify full code (not placeholders) β
β β’ Verify complete SQL migrations β
β β’ Run compatibility check against Process Map β
β β
β βΌ β
β β
β STEP 5: Update Process Map β
β ββββββββββββββββββββββββββ β
β β’ Add to Compiled Specs Registry β
β β’ Update coverage percentages β
β β’ Add new entities to registry β
β β’ Increment Process Map version β
β β
β OUTPUT β
β ββββββ β
β β’ Compiled FeatureSpec v1.x β
β β’ Updated Process Map β
β β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
The 18 Required Sections
Every compiled spec includes:
| # | Section | Purpose | Minimum Content |
|---|---|---|---|
| 1 | Header | Identity & summary | Spec ID, version, compilation summary table |
| 2 | Feature Overview | Business context | User stories, scope, key principles |
| 3 | Architecture Overview | System diagram | Layer descriptions, data flow |
| 4 | Feedback Timing Requirements | UX timing | Timing bands, haptic patterns |
| 5 | Entities | Database tables | NEW tables + EXTENSIONS with full SQL |
| 6 | Capabilities | Numbered features | CAP-### with feedback specs |
| 7 | Service Layer | Business logic | Full code with optimistic patterns |
| 8 | UI Components | Frontend code | Full components with all states |
| 9 | Database Migrations | Schema changes | Complete SQL with indexes, triggers |
| 10 | API Layer | Endpoints/functions | Production-ready, not stubs |
| 11 | Authorization | Permissions/policies | Granular, role-based |
| 12 | Observability | Monitoring | Metrics, logs, alerts with runbooks |
| 13 | Feature Flags | Rollout controls | Flags with scope and defaults |
| 14 | Test Scenarios | Quality assurance | Security, concurrency, failure, load |
| 15 | Verification Checklist | Implementation validation | Checklist for implementers |
| 16 | Implementation Priority | Phased delivery | Priority order, dependencies |
| 17 | Files to Create/Modify | Directory structure | Full file paths |
| 18 | Compilation Summary | Audit results | Senior Engineer issues & resolutions |
β See references/18-section-template.md for detailed section guidance
Minimum Requirements
template_compliance:
lines:
minimum: 1500
target: 2000+
exemplar: 3000+
sections:
required: 18
all_present: true
capabilities:
numbered: true # CAP-001, CAP-002, ...
feedback_block: true # Each has timing/haptic/visual/optimistic specs
code:
service_layer: "Full implementation, not pseudocode"
ui_components: "Full components, not wireframes"
migrations: "Complete SQL, not schema descriptions"
api_layer: "Production-ready, not stubs"
quality:
spec_id: "SPEC-### assigned"
compilation_summary: "Senior Engineer audit table with resolutions"
test_scenarios: "Minimum 3 per category"
Code Completeness Rules
Never write:
- // TODO: implement
- ... rest of implementation
- Pseudocode or placeholders
- Truncated functions
Always write:
- Full, working implementations
- All edge cases handled
- Error handling included
- TypeScript/language types complete
Capability Format
Each capability MUST have:
capability: capability_name
id: CAP-###
description: What it does
actor: Who triggers it
trigger: What triggers it
input:
field1: type
field2: type
output:
result: type
validation:
- Rule 1
- Rule 2
side_effects:
- Effect 1
- Effect 2
feedback:
timing:
input_acknowledgment: 0ms
local_render: <50ms
server_confirm: background
haptic:
on_action: light_impact
on_error: error_pattern
visual:
pending: Description
success: Description
error: Description
optimistic:
strategy: Description
rollback: Description
error_handling:
ERROR_CODE_1: "User message"
ERROR_CODE_2: "User message"
β See references/capability-format.md for detailed examples
Feedback Framework
The Paintbrush Principle
A paintbrush feels instantaneous because feedback arrives before you consciously expect it. The brain takes ~100ms to form "did that work?" β anything under 50ms feels like cause-and-effect.
Goal: Every interaction should pass the Paintbrush Test.
Timing Bands
| Band | Target | Perception | Use For |
|---|---|---|---|
| 0ms | At input event | Extension of body | Haptic, press state, selection |
| <16ms | Every frame | Fluid tracking | Cursor, drag, scroll |
| <50ms | Perceived instant | Cause-and-effect | Optimistic UI, count update |
| <150ms | Animated | Smooth transition | Dropdown, modal appear |
| <300ms | Spring settle | Physics-based | Drop into place, reorder |
| Background | Never blocks UI | Invisible | Network calls, sync |
β See references/feedback-framework.md for the Four Gaps and interaction patterns
Senior Engineer Audit
Every compiled spec includes an audit table:
| Issue | Category | Resolution |
|-------|----------|------------|
| No rate limiting | Security | Added 100 req/min limit |
| Race condition on update | Concurrency | Added optimistic locking |
| Missing index | Performance | Added composite index |
Audit Categories
| Category | What to Check |
|---|---|
| Security | Input validation, auth checks, rate limiting, injection prevention |
| Concurrency | Optimistic locking, idempotency, race conditions, deadlocks |
| Reliability | Retry logic, graceful degradation, timeouts, circuit breakers |
| Scalability | Batch processing, indexes, N+1 queries, unbounded queries |
| Observability | Metrics, structured logging, alerts, runbooks |
| Data Model | Constraints, triggers, soft delete, state machines |
β See references/senior-engineer-audit.md for full checklist
Process Map
The Process Map is the single source of truth for:
- All entities and their schemas
- All services/hooks and their purposes
- Coverage percentages per process
- Compiled specs registry
After Every Compilation
- Add new entities to Entity Registry
- Update coverage percentages
- Add to Compiled Specs Registry
- Increment version number
- Add changelog entry
β See references/process-map-template.md for blank template
Stack Adaptation
Map generic sections to your stack:
| Generic Section | React/Supabase | Django/Postgres | Rails/Postgres |
|---|---|---|---|
| Service Layer | React Hooks | Services/Managers | Service Objects |
| UI Components | TSX Components | Django Templates | ERB Views |
| API Layer | Edge Functions | Views/Viewsets | Controllers |
| Authorization | RLS Policies | Permissions | Pundit Policies |
| Realtime | Supabase Realtime | Django Channels | ActionCable |
Output Format
Header Template
# [Feature] FeatureSpec v1.1
| Property | Value |
|----------|-------|
| **Spec ID** | SPEC-### |
| **Version** | 1.1 (Compiled) |
| **Status** | Ready for Implementation |
| **Target Project** | [Project Name] |
| **Compilation Date** | [Date] |
| **Lines** | [Count] |
| **Capabilities** | [Count] |
Compilation Summary Template
## Compilation Summary
| Pass | Focus | Key Additions |
|------|-------|---------------|
| 1 | Compatibility Alignment | Entity verification, FK references |
| 2 | Feedback Framework | Timing bands, haptic patterns, optimistic UI |
| 3 | Senior Engineer Review | Security, concurrency, reliability fixes |
| 4 | Process Map Update | Coverage %, entity registry |
### Senior Engineer Audit Results
| Category | Issues Found | Issues Resolved | Status |
|----------|--------------|-----------------|--------|
| Security | X | X | β
|
| Concurrency | X | X | β
|
| Reliability | X | X | β
|
| **Total** | **Y** | **Y** | β
|
Quality Checklist
Before presenting a compiled spec:
- [ ] Spec ID assigned (SPEC-###)
- [ ] All 18 sections present
- [ ] 1,500+ lines minimum
- [ ] All capabilities numbered (CAP-###)
- [ ] All capabilities have feedback block
- [ ] All code is complete (no TODOs)
- [ ] Senior Engineer audit table filled
- [ ] Test scenarios: 3+ per category
- [ ] Process Map updated
Relationship to Other Skills
| Skill | Relationship |
|---|---|
requirements |
Requirements feed into spec compilation |
architect |
Spec includes architecture; architect provides high-level design |
frontend-design |
(Frontend systems) UI/UX section becomes DESIGN.md |
implement |
Spec is the implementation contract |
code-verification |
Verification checks against spec patterns |
code-validation |
Validation checks against spec requirements |
code-review |
Review verifies spec was implemented correctly |
Key Principles
Complete, not partial. Every section is filled. Every code block works. No placeholders.
Audited, not assumed. Senior engineer review catches issues before implementation.
Traceable, not ad-hoc. Spec IDs, capability numbers, entity references create audit trail.
Living document. Specs version. Updates increment version. Old versions are preserved.
Process Map is truth. Always check it first. Always update it after.
References
references/18-section-template.md: Detailed section-by-section guidancereferences/feedback-framework.md: Timing bands, haptics, optimistic UI patternsreferences/capability-format.md: How to write capability specificationsreferences/senior-engineer-audit.md: Full audit checklist by categoryreferences/process-map-template.md: Blank Process Map templatereferences/memory-axioms.md: Principles for maintaining project memory
# Related Skills
Use when starting any conversation - establishes how to find and use skills, requiring Skill...
|
|
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Amp
Antigravity
Claude Code
Clawdbot
Codex
Cursor
Droid
Gemini CLI
GitHub Copilot
Goose
Kilo Code
Kiro CLI
OpenCode
Roo Code
Trae
Windsurf
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.