Skip to content

Latest commit

 

History

History
499 lines (404 loc) · 21 KB

File metadata and controls

499 lines (404 loc) · 21 KB

Skills Reference — agentflow

Complete catalog of skills integrated into the governance loop. Each skill is a reusable prompt module that plugs into a specific loop stage.


Skill Dependencies & Prerequisites

Skills don't run in isolation — they depend on pipeline state and often chain together.

Dependency Map

                    ┌─────────────────┐
                    │  /sc:brainstorm  │  Ideation → Refining
                    └────────┬────────┘
                             │
                    ┌────────▼────────────────┐
                    │  /requirements-clarity   │  Catch ambiguities
                    └────────┬────────────────┘
                             │
                    ┌────────▼────────┐
                    │  /sc:spec-panel  │  Refining → Ready (gate: >= 7.0)
                    └────────┬────────┘
                             │
                    ┌────────▼────────┐
                    │   /sc:workflow   │  Ready → TODO-Today queue
                    └────────┬────────┘
                             │
                    ┌────────▼────────┐
                    │  /sh:ux-design  │  (if user flow exists)
                    │  wireframe →    │  validate → graduate
                    │  /frontend-     │  → hi-fi on top
                    │   design        │
                    └────────┬────────┘
                             │
              ┌──────────────▼──────────────┐
              │        INNER LOOP           │
              │                             │
              │  ┌───────────────────────┐  │
              │  │ /test-driven-dev      │  │  Step 6: Write test first
              │  └──────────┬────────────┘  │
              │             ▼               │
              │  ┌───────────────────────┐  │
              │  │ /verification         │  │  Step 7: Verify ACs
              │  └──────────┬────────────┘  │
              │             ▼               │
              │  ┌───────────────────────┐  │
              │  │ /requesting-review    │  │  Step 8: Self-review
              │  │ /receiving-review     │  │
              │  └──────────┬────────────┘  │
              │             ▼               │
              │  ┌───────────────────────┐  │
              │  │ CLEANUP SUB-LOOP      │  │  Step 9: Quality gate
              │  │  /sc:analyze          │  │
              │  │  /production-code-audit│ │  (M+ tasks only)
              │  │  /sc:cleanup          │  │  (enforces /clean-code)
              │  └──────────┬────────────┘  │
              │             ▼               │
              │  ┌───────────────────────┐  │
              │  │ /commit-smart         │  │  Step 10: Atomic commit
              │  └──────────┬────────────┘  │
              │             ▼               │
              │  ┌───────────────────────┐  │
              │  │ /finishing-branch     │  │  Step 11: PR + merge
              │  └──────────┬────────────┘  │
              │             ▼               │
              │  ┌───────────────────────┐  │
              │  │ /session-handoff      │  │  Step 13: Save context
              │  └───────────────────────┘  │
              └─────────────────────────────┘
                             │
                      every 10 stories
                             │
                    ┌────────▼────────┐
                    │    /kaizen      │  Retrospective
                    └─────────────────┘

Prerequisites per Skill

Skill Requires Produces
/sc:brainstorm Raw idea in INBOX or BACKLOG#Ideation Requirements document
/requirements-clarity Brainstorm output exists Ambiguity report, clarified requirements
/sc:spec-panel Spec document with US + AC Score (gate: >= 7.0), improvement suggestions
/sh:ux-design User flow (Figma/Miro/PRD), optional business-panel brief Clickable wireframe prototype, component mapping, handoff doc
/sc:design Requirements doc (architecture tasks) Architecture specification
/sc:workflow Ready items in BACKLOG Populated TODO-Today.md queue
/test-driven-development Task in queue with clear AC Failing test file
/verification-before-completion Implementation complete, tests passing Evidence log (test output, screenshots)
/requesting-code-review Changed files in working tree Review findings
/receiving-code-review Review feedback received Applied fixes
/sc:analyze Implementation complete Quality findings (Low/Medium/High)
/production-code-audit Task size >= M (Medium) Security/perf/arch findings
/sc:cleanup Findings from /sc:analyze Fixed code, clean findings
/commit-smart Clean greenlight, all findings resolved Atomic commit
/finishing-a-development-branch On a feature branch, commit done PR created, branch strategy decided
/session-handoff Task(s) complete or session ending HANDOVER.md with resume checklist
/kaizen 10 User Stories completed since last retro Actionable improvement items

Skill Chains

Common sequences of skills that execute together as workflows.

Chain 1: Graduation Pipeline

Trigger: New idea needs to become implementable work.

/sc:brainstorm "idea description" --depth deep
    ↓ produces requirements doc
/requirements-clarity
    ↓ catches ambiguities, iterates
/sc:spec-panel requirements/SPEC.md --mode critique --focus requirements
    ↓ gate: score >= 7.0 (else iterate)
/sc:workflow requirements/SPEC.md --strategy systematic
    ↓ populates TODO-Today.md queue

Example:

User sends: "We need dark mode support"

1. /sc:brainstorm "dark mode support" --depth deep
   → Outputs requirements doc with scope, edge cases, themes

2. /requirements-clarity
   → Catches: "Does dark mode apply to exported assets or just UI?"
   → Clarifies: UI only, export uses explicit color schemes

3. /sc:spec-panel requirements/SPEC_DARK_MODE.md --mode critique
   → Score: 7.5/10 — Ready

4. /sc:workflow requirements/SPEC_DARK_MODE.md --strategy systematic
   → Generates 5 queue items in TODO-Today.md

Chain 1b: UX Wireframe Pipeline

Trigger: Feature has a multi-step user flow (onboarding, checkout, wizard, form).

/sh:business-panel "feature PRD" --focus competitive
    ↓ strategic constraints (JTBD, ICP, risk flags)
/sh:ux-design <figma-url or PRD>
    ↓ Phase 1-3: ingest flow → plan → generate wireframe
    ↓ Phase 4-5: concept branching → compare & decide
    ↓ Phase 6-8: iterate → graduation gate
/frontend-design
    ↓ builds hi-fi ON TOP of wireframe (no restart)
/sh:plan
    ↓ implementation plan for remaining backend/integration work

Example:

User sends: "We need an onboarding wizard for new users"

1. /sh:business-panel requirements/PRD_ONBOARDING.md --focus growth
   → Christensen: JTBD = "get first meeting scheduled within 5 min"
   → Godin: "reduce steps to under 5 — each step is a drop-off cliff"
   → Taleb: "if OAuth fails, user is stuck — mandatory fallback"

2. /sh:ux-design figma.com/board/abc123
   → Generates clickable 5-step wireframe in React (target stack)
   → Playwright smoke test: all steps clickable, no dead ends
   → User: "Give me option B as chat-based onboarding"
   → Concept B generated, comparison table, user picks chat-based

3. /frontend-design
   → Replaces grayscale with design tokens, placeholders with components
   → Preserves flow structure and navigation from wireframe

4. /sh:plan
   → Plans backend API, auth integration, analytics events

Chain 2: Implementation Cycle (per task)

Trigger: First unchecked - [ ] item in TODO-Today.md.

/test-driven-development
    ↓ write failing test
[implement until test passes]
    ↓
/verification-before-completion
    ↓ verify ACs with evidence
/requesting-code-review
    ↓ self-review changed files
/receiving-code-review (if issues found)
    ↓ apply fixes

Example:

Task: "- [ ] Implement US-DM-01: CSS custom property theme tokens"

1. /test-driven-development
   → Creates tests/test_dark_mode_tokens.py
   → Tests assert :root has --bg-primary, --text-primary, etc.
   → Tests fail (tokens don't exist yet)

2. [Implementation]
   → Add CSS custom properties to shell.css
   → Tests pass

3. /verification-before-completion
   → Evidence: test output (5/5 pass), screenshot of both themes

4. /requesting-code-review
   → Reviews shell.css changes
   → Finding: "hardcoded fallback #fff should use var(--bg-default)"

5. /receiving-code-review
   → Evaluates finding: valid, applies fix

Chain 3: Quality Tail (mandatory, end of every batch)

Trigger: Implementation complete, before commit.

/sc:analyze "<changed files>" --focus quality
    ↓ identifies findings (Low/Medium/High)
/production-code-audit (if task size >= M)
    ↓ deep security/perf/architecture scan
/sc:cleanup --type all
    ↓ fixes Low findings, enforces /clean-code
[if Medium+ findings: STOP — human review]
    ↓
/commit-smart
    ↓ atomic conventional commit
/finishing-a-development-branch (if on feature branch)
    ↓ PR creation, merge strategy

Example:

Changed files: app/services/theme_service.py, app/static/css/shell.css

1. /sc:analyze "app/services/theme_service.py app/static/css/shell.css"
   → 2 Low findings: missing type hint, unused import

2. /production-code-audit (task was Medium)
   → No security issues, no perf concerns

3. /sc:cleanup --type all
   → Fixes: adds type hint, removes unused import
   → Re-runs greenlight: all green

4. /commit-smart
   → "feat(theme): add CSS custom property tokens for dark mode"

5. /finishing-a-development-branch
   → Creates PR: "Dark mode: CSS token system"
   → Suggests squash merge

Chain 4: Context Management

Trigger: Session ending or context window approaching limits.

/session-handoff
    ↓ produces HANDOVER.md with:
    ↓   - git state (branch, modified files)
    ↓   - decisions made this session
    ↓   - open questions
    ↓   - resume checklist for next session

Chain 5: Retrospective

Trigger: 10 User Stories completed (counter in project memory).

/kaizen
    ↓ guides continuous improvement:
    ↓   - root cause analysis of recent issues
    ↓   - error proofing opportunities
    ↓   - standardization candidates
    ↓   - actionable items → BACKLOG#Ideation or CLAUDE.md rule updates

Skills by Category

Dev Loop Skills (wired to inner loop steps)

Skill Step What It Does When to Skip
/test-driven-development 6 Writes failing test before implementation Non-US tasks (spikes, docs)
/verification-before-completion 7 Verifies acceptance criteria with evidence Never — always required
/requesting-code-review 8 Self-reviews changed files for issues Trivial 1-line fixes
/receiving-code-review 8 Evaluates and applies review feedback No feedback to process
/production-code-audit 9 Deep security/perf/arch scan Tasks sized S (Small)
/clean-code 9 Coding standards during cleanup fixes N/A (enforced by /sc:cleanup)
/commit-smart 10 Semantic conventional commit with context Never — always required
/finishing-a-development-branch 11 PR creation, branch cleanup, merge Not on a feature branch
/session-handoff 13 Context save for next session Never — always at session end
/kaizen retro Continuous improvement retrospective Counter < 10 stories

Support Process Skills (parallel to dev loop)

SP-1: Go-to-Market Pipeline

Trigger: New product/feature release, marketplace listing. Cadence: Per release. Owner: Human initiates, agent executes.

Step Skill Purpose
1 /marketing-strategy-pmm Positioning, ICP definition, messaging framework
2 /launch-strategy Launch plan, timeline, channel selection
3+ /executing-marketing-campaigns Campaign execution across channels
3+ /copywriting Landing page, listing, ad copy
3+ /copy-editing Polish and refine existing copy
3+ /social-content Platform-specific social media content
3+ /marketing-ideas Brainstorm campaign concepts

Feedback loop: Campaign metrics → BACKLOG#Ideation for product improvements.


SP-2: SEO & Discovery Engine

Trigger: Monthly audit cycle or new content published. Cadence: Monthly audit + per-listing optimization. Owner: Human initiates audit, agent runs optimization.

Step Skill Purpose
1 /seo-audit Technical SEO health check
2 /seo-optimizer Content and keyword optimization
2 /programmatic-seo Template-driven pages at scale
2 /schema-markup Structured data for rich results
2 /analytics-tracking Event tracking setup
3 /google-analytics Traffic analysis and insights
3 /roier-seo Lighthouse/PageSpeed performance audit

Feedback loop: SEO findings → BACKLOG#Ideation for technical fixes.


SP-3: Competitive Intelligence Radar

Trigger: Planning round, new competitor detected, quarterly review. Cadence: Per planning round or quarterly. Owner: Human initiates, agent researches.

Step Skill Purpose
1 /competitor-alternatives Competitor comparison and mapping
2 /pricing-strategy Pricing model analysis and recommendations
3 /free-tool-strategy Free tool as marketing/acquisition lever
4 /x-twitter-scraper Social signal and trend monitoring
5 /app-store-optimization Marketplace listing optimization

Feedback loop: Competitive insights → BACKLOG priorities, pricing adjustments.


SP-4: Content Production Desk

Trigger: New listing needed, content calendar slot. Cadence: Per listing or weekly. Owner: Human assigns topic, agent produces.

Step Skill Purpose
1 /content-research-writer Research-backed articles with citations
2 /content-creator SEO-optimized marketing content
3 /copy-editing Final polish pass
4 /viral-generator-builder Shareable interactive tools
5 /shopify-development E-commerce integration

Feedback loop: Content performance → BACKLOG#Ideation for new topics.


SP-5: Lightsout Hygiene

Trigger: End-of-day (queue empty + autopilot=run), planning round agenda item, or on-demand. Cadence: Daily (if autopilot active) or weekly minimum. Owner: Autopilot-triggered or human-initiated.

/health                        → pipeline staleness + velocity snapshot
/backlog-grooming              → archive shipped items, flag stale/orphans
/session-handoff               → save context for next session
Step Skill Purpose
1 /health Pipeline staleness scan, velocity trend, risk review
2 /backlog-grooming Archive shipped items, detect stale/orphans, cleanup fixes
3 /session-handoff Save decisions, open questions, resume checklist

Workflow:

  1. /health produces a read-only pipeline snapshot
  2. /backlog-grooming runs the archive + cleanup pass (if candidates found)
  3. /session-handoff captures what was cleaned and any open decisions

Trigger conditions (any one):

  • Autopilot queue is empty (natural end-of-day)
  • Planning round (PO-05 agenda item #2)
  • User invokes /backlog-grooming directly
  • Weekly timer fires (if using /loop)

Feedback loop: Stale items → user decides: graduate, archive, or add Next action. Orphans → BACKLOG cleanup or deletion.


On-Demand Toolbox (invoke as needed)

Skill Category Use When
/sh:ux-design Design Clickable wireframe prototypes — validate flow before hi-fi
/figma Design Translating Figma designs into code
/frontend-design Design Building production-grade UI from scratch (or graduating a wireframe)
/canvas-design Design Creating visual artifacts (PNG, PDF)
/imagegen Generation Generating or editing images via AI
/algorithmic-art Generation Creating generative art with p5.js
/excalidraw-diagram Visualization Visualizing workflows, architectures, data flows
/theme-factory Styling Applying themed templates to artifacts
/git-pushing Git Safe push with pre-flight checks
/using-git-worktrees Git Isolated feature work in git worktrees
/git-context-controller Git Managing agent memory as versioned files
/web-performance-optimization Performance Core Web Vitals, bundle size, caching

Utility Skills (available but not governance-wired)

These skills are useful but don't have a fixed position in the loop. Invoke them when the context calls for it.

Skill Purpose
/changelog-generator Auto-generate changelogs from git history
/clean-code Pragmatic coding standards reference
/code-review-checklist Comprehensive review checklist template
/crafting-effective-readmes README authoring by audience type
/create-plan Concise implementation plan for coding tasks
/doc-coauthoring Structured documentation co-authoring workflow
/executing-plans Execute written plans with review checkpoints
/planning-with-files Manus-style file-based planning (task_plan.md, findings.md)
/python-patterns Python development principles and decision-making
/reducing-entropy Minimize total codebase size (manual activation only)
/security-best-practices Language/framework-specific security review
/security-threat-model Repository-grounded threat modeling
/ship-learn-next Transform learning content into implementation plans
/skill-creator Guide for creating new skills
/software-architecture Python/FastAPI architecture patterns
/writing-plans Plan implementation strategy before coding
/writing-rules Create hookify rules for automation
/sh:poster-analysis Generate DIN A1 business analysis poster as PDF

Quick Reference: "Which skill do I need?"

I want to... Use this skill
Turn a raw idea into a spec /sc:brainstorm/requirements-clarity/sc:spec-panel
Start implementing a feature /test-driven-development
Check if my work is really done /verification-before-completion
Review my own code /requesting-code-review
Respond to review feedback /receiving-code-review
Run a quality check /sc:analyze/sc:cleanup
Do a deep security/perf audit /production-code-audit
Make a good commit /commit-smart
Create a PR and clean up branch /finishing-a-development-branch
Save context for next session /session-handoff
Run a team retrospective /kaizen
Write marketing copy /copywriting or /copy-editing
Optimize for search engines /seo-audit/seo-optimizer
Analyze competitors /competitor-alternatives/pricing-strategy
Create visual content /canvas-design or /imagegen
Create a business analysis poster /sh:poster-analysis
Wireframe a user flow /sh:ux-design/frontend-design
Build a UI component /frontend-design or /figma
Plan before coding /create-plan or /writing-plans
Create a new skill /skill-creator

61 skills wired to governance. 18 utility skills available on demand. Zero orphans.