Skip to content

Skills Index

Audience: AI Agents + Humans (canonical skill listing)

This document lists all available skills and how to activate them.

Machine-readable source: manifest.yaml — canonical skill metadata (YAML).

What Are Skills?

Skills are specialized knowledge packages that provide: - Guidelines - Principles and best practices - Workflows - Step-by-step procedures - Examples - Concrete case studies and samples - Templates - Reusable document formats - Checklists - Quick reference for reviews

Skill Tiers

Core Skills (Always Active)

These skills are automatically loaded for every project:

Skill Description Path
git-workflow Git best practices, branching strategies, commit hygiene core/git-workflow/
todo-workflow Invokable autonomous/semi-autonomous development workflow core/todo-workflow/

Optional Skills (Opt-In)

Enable these by checking them in your project's AGENTS.md:

Skill Description Path
codebase-analysis Base analysis engine producing structured analysis model optional/codebase-analysis/
analysis-outputs Output adapters (architecture-docs, coding-context, etc.) optional/analysis-outputs/
arch-analysis Architecture documentation (codebase-analysis + architecture-docs) optional/arch-analysis/
security-analysis Security posture assessment with dual output (human-readable + compliance) optional/security-analysis/
nonfunctional-analysis Testing coverage, configuration audit, performance, code health optional/nonfunctional-analysis/
architecture-synthesis Middle-out architecture from diagrams and specs optional/architecture-synthesis/
fitness-functions Evolutionary architecture fitness measurement and validation optional/fitness-functions/
structurizr C4 architecture modeling with Structurizr DSL optional/structurizr/
togaf Enterprise architecture framework (ADM phases A-H) optional/togaf/
software-design Design principles, patterns, architectural decisions optional/software-design/
tech-stack-decisions Technology evaluation, decision frameworks, ADRs optional/tech-stack-decisions/
code-conventions Style guides, naming conventions, structure standards optional/code-conventions/
presentation Generate slide decks (PPTX, PDF) from Markdown using Marp CLI optional/presentation/

How to Activate Optional Skills

In your project's root AGENTS.md, find the "Enabled Skills" section and check the skills you want:

## Enabled Skills

- [x] `software-design` - Design principles and patterns
- [x] `tech-stack-decisions` - Technology evaluation frameworks
- [ ] `code-conventions` - Style guides and conventions

When a skill is enabled, AI agents should read all files in that skill's directory before starting relevant work.


Loading Strategy

Lazy loading: Do NOT read all skill files at session start. Only load a skill when invoking it.

  • Session start: Read this index file (_index.md) for skill discovery
  • On invocation: Read the skill's README.md + workflows.md
  • As needed: Read templates.md, examples.md, checklist.md during skill execution

Token Budget

Approximate context window cost when using this toolkit.

Session Start (~4,600 tokens)

File Tokens
Toolkit AGENTS.md ~500
core/instructions.md ~440
core/glossary.md ~1,580
skills/_index.md (this file) ~2,100
Total ~4,620

Additional core docs (read when needed): core/workflows.md (~1,940), core/architecture-thinking.md (~3,170).

Per-Skill Cost (on invocation)

Skill Tokens Skill Tokens
arch-analysis ~20,100 security-analysis ~20,100
architecture-synthesis ~19,600 fitness-functions ~18,100
structurizr ~15,600 nonfunctional-analysis ~12,700
code-conventions ~8,500 codebase-analysis ~8,400
software-design ~7,300 tech-stack-decisions ~6,900
analysis-outputs ~3,900 togaf (index only) ~2,300
TOGAF phases ~10K-24K each TOGAF all phases ~154,500

Context Window Planning

Window Size Budget After Session Start Practical Limit
100K tokens ~95K for skills + code 1-2 large skills + codebase
200K tokens ~195K for skills + code 3-4 large skills + codebase

Rule of thumb: Reserve at least 50% of context for actual codebase content. A single analysis skill (arch, security, nonfunctional) uses ~15-20K tokens.


Skill File Structure

Each skill contains these files:

File Purpose When to Read
README.md Guidelines and principles On skill invocation
workflows.md Step-by-step procedures On skill invocation
examples.md Concrete examples During execution (as needed)
templates.md Reusable templates During execution (as needed)
checklist.md Quick reference During execution (as needed)

Default Output Directories

When skills generate output files, they use these defaults (confirm with user before generating):

Output Type Default Directory
Analysis analysis/
Presentations presentations/

Creating New Skills

To add a new skill:

  1. Create a directory in skills/optional/ (or skills/core/ if always-on)
  2. Add the required files following the structure above
  3. Update this index with the new skill
  4. Submit a pull request

Skill Naming Conventions

  • Use kebab-case for directory names
  • Keep names concise but descriptive
  • Avoid generic names like "best-practices"

Good Examples

  • api-design - REST/GraphQL API design principles
  • testing-strategy - Test pyramid, coverage, TDD
  • security-practices - OWASP, auth, secrets management

Avoid

  • misc - Too vague
  • stuff - Not descriptive
  • my-skill - Not meaningful

Skill Dependencies

Some skills may recommend reading other skills first:

Skill Recommended Prerequisites
git-workflow None
todo-workflow git-workflow (uses git operations)
codebase-analysis None (base analysis engine)
analysis-outputs codebase-analysis (consumes analysis model)
arch-analysis None (uses codebase-analysis internally)
security-analysis arch-analysis or codebase-analysis (recommended)
nonfunctional-analysis codebase-analysis (optional, for context)
architecture-synthesis None (parses diagrams/specs directly)
fitness-functions None (standalone or with migration-planning)
structurizr None (standalone C4 modeling)
togaf arch-analysis (recommended for baseline)
software-design None
tech-stack-decisions software-design (optional)
code-conventions None
presentation None (standalone, uses exported diagrams)

Invokable Skills

Some skills can be explicitly invoked during a session:

Skill Invocation Description
skill-discovery "What skills do you have?" / "List your capabilities" List all skills with offer to elaborate
todo-workflow "Use todo workflow" Start autonomous/semi-autonomous development
codebase-analysis "Analyze this codebase" Base analysis with multi-output adapter selection
arch-analysis "Analyze the architecture" Architecture documentation (shortcut)
security-analysis "Analyze security" Security posture assessment with compliance reports
nonfunctional-analysis "Analyze code quality" Testing, config, performance, health assessment
architecture-synthesis "Synthesize architecture from diagrams" Parse diagrams/specs into architecture model
fitness-functions "Define fitness functions" / "Analyze system fitness" Evolutionary architecture fitness assessment
structurizr "Create C4 model" Architecture modeling with Structurizr DSL
togaf "Apply TOGAF" / "Create architecture vision" Enterprise architecture using ADM
presentation "Generate presentation" / "Create slides" Markdown to PPTX/PDF slide generation

See the skill's README for full invocation options. Skill Discovery is defined in core/workflows.md.


Agent Notes

Analysis Workflow

  • Use codebase-analysis as the base engine
  • Select output adapters based on user needs
  • Run analysis once, output multiple formats

TOGAF Integration

  • Core concepts in core/architecture-thinking.md apply automatically
  • ADM phases in togaf/ are invokable as needed
  • Supports lightweight/partial use (not all-or-nothing)

Future Skills (Ideas)

Aspirational ideas — not yet planned or specified:

  • testing-strategy - Test pyramid, coverage, TDD/BDD
  • api-design - REST, GraphQL, versioning
  • documentation - Code docs, user docs, architecture docs
  • performance - Profiling, optimization, caching
  • devops - CI/CD, deployment, monitoring

See ROADMAP-TRACKER.md for planned work with specs and timelines.