
Swain Help
- 110 installs
- 2 repo stars
- Updated July 24, 2026
- cristoslc/swain
Invoke when working in a Swain-powered repo and you need command usage, workflow steps, configuration options, or troubleshooting guidance without leaving the agent session.
About
swain-help is a documentation-oriented Claude Code skill from cristoslc/swain that lets agents fetch Swain-specific usage, commands, configuration, and troubleshooting guidance while building agent or CLI workflows, keeping implementation aligned with the project's intended Swain patterns.
- In-repo Swain command and workflow reference
- Reduces context switching to external README hunting
- Agent-friendly help retrieval during implementation
- Covers configuration and troubleshooting paths
- Pairs with Swain automation in Claude Code sessions
Swain Help by the numbers
- 110 all-time installs (skills.sh)
- Ranked #634 of 1,879 Documentation skills by installs in the Skillselion catalog
- Data as of Jul 29, 2026 (Skillselion catalog sync)
npx skills add https://github.com/cristoslc/swain --skill swain-helpAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Installs | 110 |
|---|---|
| repo stars | ★ 2 |
| Last updated | July 24, 2026 |
| Repository | cristoslc/swain ↗ |
What it does
Invoke when working in a Swain-powered repo and you need command usage, workflow steps, configuration options, or troubleshooting guidance without leaving the agent session.
Files
<!-- swain-model-hint: sonnet, effort: medium — default for conceptual explanations; see per-section overrides below -->
swain-help
Contextual help for the swain skill ecosystem.
Mode detection
Determine the mode from context:
| Signal | Mode |
|---|---|
| Invoked from swain-init Phase 4, or user says "just set up swain" / "what now after init" | Onboarding |
| User asks a specific question ("how do I...", "what is...", "when should I...") | Question |
| User asks for a reference, cheat sheet, commands, or overview | Reference |
Onboarding mode
Present a concise orientation — help the user understand what they just installed without overwhelming them. Adapt tone to context (first-time dev vs experienced engineer).
Present this:
Welcome to swain. Here's how it works:
>
The big picture: Swain manages your project's documentation artifacts (specs, epics, ADRs, etc.) and tracks implementation work — so nothing falls through the cracks between sessions.
>
Three things to know:
>
1. `/swain` is your entry point. It routes to the right sub-skill automatically. You can also call skills directly (/swain-design,/swain-do, etc.).
>
2. Design before you build. When you want to implement something, start with /swain to create a spec. Swain enforces a "plan before code" workflow — it creates tracked tasks before implementation begins.>
3. Health checks are automatic./swain-doctorruns at session start to ensure routing rules are in place and.tickets/is healthy. You don't need to think about it.
>
Common starting points:
- "I want to plan a new feature" → creates an Epic or Spec
- "Write a spec for X" → creates an Agent Spec
- "What should I work on next?" → checks your task backlog
- "File a bug" → creates a Spec with type: bug- "Let's release" → version bump + changelog
>
Need more? Ask me anything about swain, or say /swain help reference for a full cheat sheet.Then stop. Let the user ask follow-up questions — don't dump everything at once.
Question mode
Answer the user's specific question using your knowledge of swain. If you need details beyond what's in this skill, read the relevant reference:
| Topic | Where to look |
|---|---|
| Artifact types, phases, relationships | references/quick-ref.md — Artifacts section |
| Commands and invocations | references/quick-ref.md — Commands section |
| Step-by-step walkthroughs | references/workflows.md |
| Artifact definitions and templates | sibling skill swain-design/references/<type>-definition.md |
| tk (ticket) CLI reference | sibling skill swain-do/references/tk-cheatsheet.md |
| Troubleshooting | sibling skill swain-design/references/troubleshooting.md |
Guidelines for answering:
- Be concise. Answer the question, don't dump the entire reference.
- Use examples when they clarify — "You'd say
/swain create a spec for auth token rotation". - Hand off when appropriate. If the user's question is really a request to do something (e.g., "how do I create a spec?" followed by "ok do it"), invoke the relevant skill directly via the Skill tool. Explain what you're doing: "I'll hand this off to swain-design."
- Admit gaps. If something isn't covered, say so rather than inventing swain features.
<!-- swain-model-hint: haiku, effort: low — reference lookups are simple file reads -->
Reference mode
When the user wants an overview or cheat sheet, read references/quick-ref.md and present the relevant section. If they want "everything", present the full quick reference but note it's dense.
For workflow walkthroughs, read references/workflows.md.
Swain Quick Reference
Skills at a Glance
| Skill | Invoke with | What it does |
|---|---|---|
| swain | /swain <request> | Routes to the right sub-skill |
| swain-init | /swain init | One-time project setup |
| swain-doctor | /swain-doctor | Session-start health checks (automatic) |
| swain-session | /swain-session or "what's next?" | Context bookmarks, preferences, dashboard — active work, blockers, next steps, GitHub issues |
| swain-design | /swain-design or /swain + artifact request | Create and manage documentation artifacts |
| swain-search | /swain-search or /swain + research request | Collect and cache evidence pools |
| swain-do | /swain-do or /swain + task request | Track tasks and implementation work |
| swain-sync | /swain-sync or /swain sync | Fetch, rebase, commit, and push changes |
| swain-release | /swain-release or /swain release | Version bump, changelog, git tag |
| swain-keys | /swain-keys | Per-project SSH keys for git signing and auth |
| swain-update | /swain-update or /swain update | Update swain to latest version |
| swain-retro | /swain-retro or /swain retro | Capture learnings at EPIC completion or on demand |
| swain-help | /swain help or /swain-help | This help system |
Artifacts
Swain manages 10 artifact types, organized into three lifecycle tracks.
Implementable track (tracked via tk)
| Type | ID Pattern | Phases | When to use |
|---|---|---|---|
| Agent Spec | SPEC-NNN | Proposed → Ready → In Progress → Needs Manual Test → Complete | Technical specification for an agent or component. Supports `type: enhancement \ |
These require a tracked plan (via swain-do) before implementation begins.
Container track (children are tracked)
| Type | ID Pattern | Phases | When to use |
|---|---|---|---|
| Initiative | INITIATIVE-NNN | Proposed → Active → Complete | Strategic grouping of Epics under a Vision — prioritization and decision tracking |
| Epic | EPIC-NNN | Proposed → Active → Complete | Large deliverable decomposed into specs |
| Spike | SPIKE-NNN | Proposed → Active → Complete | Time-boxed investigation to reduce uncertainty |
Standing track (no tracking)
| Type | ID Pattern | When to use |
|---|---|---|
| Vision | VISION-NNN | Product direction and goals |
| Journey | JOURNEY-NNN | User journey with pain points |
| ADR | ADR-NNN | Architectural decision record |
| Persona | PERSONA-NNN | User persona definition |
| Runbook | RUNBOOK-NNN | Operational procedure |
| Design | DESIGN-NNN | UI/UX design artifact |
Artifact relationships
- Vision → decomposes into Initiatives, Epics, and Journeys
- Initiative → groups related Epics (and optionally Specs) under a Vision
- Epic → decomposes into Specs, Spikes
- Spec → may reference ADRs, Personas, Designs
- Spike → attaches to any artifact, may produce ADRs
- Any artifact can declare
depends-on:blocking dependencies
When to use which
- Initiative vs Epic: Initiative = strategic direction with multiple deliverables ("harden security"). Epic = single deliverable with multiple specs ("build scanning tool"). If it needs 2+ epics, it's an Initiative.
- Spec under Initiative: Small work (bugs, enhancements) can attach directly to an Initiative without an Epic wrapper. If it clusters, promote to an Epic.
- Spec under Epic vs standalone Spec: Use
parent-epicwhen the spec is part of a planned deliverable. Standalone specs are for one-off fixes.
Commands
Creating artifacts
/swain create a vision for X
/swain write a spec for Y
/swain file a bug about Z
/swain plan an epic for W
/swain create an ADR for this decision
/swain create a runbook for deploymentManaging lifecycle
/swain move SPEC-001 to Ready
/swain transition SPEC-003 to Complete
/swain abandon SPIKE-002Task tracking
/swain what should I work on next?
/swain show my tasks
/swain create a plan for SPEC-001Validation and auditing
/swain check for stale references
/swain show the dependency graph
/swain validate ADRsReleasing and committing
/swain sync
/swain release
/swain bump versionKey Concepts
The "plan before code" rule
When a SPEC comes up for implementation, swain requires a tracked plan (via tk) before code is written. This ensures work is visible and manageable across sessions. Swain-design enforces this automatically — when you transition an artifact to its implementation phase, it triggers swain-do to create the plan.
tk (ticket)
The vendored, git-backed task tracker swain uses. Verified by swain-init, operated by swain-do. Key commands:
| Command | What it does |
|---|---|
tk ready | Show next task to work on (blocker-aware) |
tk create "title" -t task | Create a task |
tk claim <id> | Claim work |
tk close <id> | Mark complete |
tk ready | Overview of ready work |
tk blocked | Show blocked tasks |
Governance block
The <!-- swain governance --> block in AGENTS.md contains routing rules that make swain skills discoverable. Managed automatically by swain-doctor. Don't edit it manually — customize anything outside the markers.
The @AGENTS.md pattern
CLAUDE.md contains just @AGENTS.md, which includes the full AGENTS.md file. This lets one file serve Claude Code, GitHub, Cursor, and other tools that read AGENTS.md natively.
Project Structure
<project>/
├── CLAUDE.md # Contains: @AGENTS.md
├── AGENTS.md # Project instructions + governance block
├── .tickets/ # tk database (git-tracked)
├── .agents/ # Swain config and logs
└── docs/
├── vision/ # VISION artifacts
├── initiative/ # INITIATIVE artifacts
├── epic/ # EPIC artifacts
├── spec/ # SPEC artifacts
├── spike/ # SPIKE artifacts
├── adr/ # ADR artifacts
├── persona/ # PERSONA artifacts
├── runbook/ # RUNBOOK artifacts
├── design/ # DESIGN artifacts
├── journey/ # JOURNEY artifacts
└── list-*.md # Lifecycle indexes per typeCommon Workflows
New feature (end-to-end)
1. Define scope: /swain create an epic for user authentication
- Swain-design creates EPIC-NNN with scope, success criteria, and acceptance tests
2. Decompose: /swain create a spec for JWT token handling (reference the epic)
- Creates SPEC-NNN linked to the epic
3. Plan implementation: When the spec reaches Ready, swain-design triggers swain-do to create a tracked plan with tasks
4. Work the plan: /swain what should I work on?
- Swain-do shows the next ready task (blocker-aware)
- Claim it, do the work, mark complete
5. Commit: /swain sync
6. Release: /swain release when the epic is complete
Bug fix
1. File the bug: /swain file a bug: login fails when password contains special characters
- Creates a SPEC with
type: bug, including reproduction steps, severity, and expected vs. actual behavior
2. Plan the fix: Swain creates tracked tasks before code changes begin
3. Fix and verify: Work the tasks, mark resolved, verify
4. Commit: /swain sync
Research spike
1. Create the spike: /swain create a spike to evaluate WebSocket vs SSE for real-time updates
- Time-boxed investigation with clear questions to answer
2. Do the research: Spike moves to Active
3. Record findings: Complete the spike with conclusions
- May produce an ADR if an architectural decision was made
Checking project status
1. Get the dashboard: /swain-session or just ask "what's next?" / "where are we?"
- Shows active epics with progress ratios (e.g., 3/5 specs complete)
- Surfaces blocked items, in-progress tasks, and GitHub issues
- Provides a ranked recommendation for what to work on next
2. Drill into specifics: Follow up on anything in the dashboard
- "Tell me more about EPIC-003"
- "What's blocking SPEC-012?"
3. Act on the recommendation: The dashboard points you to the highest-leverage next step
Starting a new session
1. Health check: /swain-doctor runs automatically (or invoke manually)
2. Context restore: /swain-session runs automatically — restores your last context bookmark
3. See what's in progress: /swain-session or tk ready
4. Pick up work: Follow the dashboard's recommendation or ask /swain what should I work on?
Adopting swain in an existing project
1. Run init: /swain init
- Migrates CLAUDE.md, verifies tk, adds governance rules
2. Orientation: Swain-help walks you through what's available
3. Start creating artifacts: /swain create a vision for this project
Artifact lifecycle walkthrough
A typical implementation-tier artifact (Spec) goes through:
1. Proposed — artifact lands in docs/spec/Proposed/ 2. Ready — transition when scope, acceptance criteria, and dependencies are confirmed 3. In Progress — swain-do creates tracked tasks; implementation begins 4. Needs Manual Test — all tasks complete; populate the Verification table with evidence 5. Complete — artifact moves to docs/spec/Complete/ after all criteria pass 6. Validate — specwatch checks for stale refs; adr-check validates compliance