By name: tld-help description: | Quick reference for all TLD skills. Shows what each skill does, when to use it, and the standard flow. Use this skill whenever the user says "tld-help", "tld help", "what are my options", "what can I run", "help", or needs a reminder of the available TLD commands.
TLD Help
Print the reference card below, then determine the user's current position in the flow and present context-aware options.
TLD Skills Reference
Core Flow (run in order)
| # | Skill | What it does | When to use |
|---|---|---|---|
| 1 | /tld-setup |
Loads next ticket from milestone + Linear | Start of every ticket |
| 2 | /tld-write-tests |
Writes failing tests from the AC | After setup (red phase) |
| 3 | /tld-build |
Implements code to make tests pass | After tests are written (green phase) |
| 4 | /tld-audit |
Security + architecture review | After build, before verify (optional) |
| 5 | /tld-run-test |
Runs tests, drift check, manual QA, commits on approval | After build or audit |
| 6 | /tld-next |
Marks ticket Done, determines next step · may mutate next-ticket labels via override loop | After successful commit |
Automation
Reading the names: the prefix says who lands the commit — tld-* = you control commit/push/PR (hands-on, code tickets), npc-* = the skill commits for you (hands-off, content/doc tickets). The suffix says how many human stops — partial = more stops, full = fewest stops the family's safety allows.
| Skill | What it does | When to use |
|---|---|---|
/tld-partial-auto |
Chains the full pipeline with 2 human gates (test-spec review + QA), commits on approval, marks Done | Code tickets you want automated but with review checkpoints |
/tld-full-auto |
Runs the pipeline hands-off to a verified checkpoint, then STOPS before commit and preps your manual check; only flags real problems | Code tickets you want driven to ready-to-land unattended, keeping the commit/PR in your hands |
/tld-pr |
Lands a verified ticket: commit → push → open PR, then stops before merge | After /tld-full-auto's checkpoint (or any committed ticket) when you're ready to ship |
/npc-partial |
Build → STOP for manual QA on uncommitted diff → commit + tld-next on approval | Doc/content tickets where test command is skip and you want one QA pause |
/npc-full |
Build → commit → tld-next, no review pause | Doc/content tickets where test command is skip and you trust the build |
Recovery + Navigation
| Skill | What it does | When to use |
|---|---|---|
/tld-align |
Fixes implementation after test failures | After /tld-run-test fails |
/tld-commit |
Commits the current ticket; asks commit only (stay In Progress) vs commit and progress (mark Done + next). No push/PR. | Per-ticket landing in a multi-ticket story; or finishing a commit after a detour |
/tld-pr |
Lands a verified ticket: commit → push → open PR, stops before merge | Story end — one PR for all the story's tickets |
/tld-skip |
Reverts to Todo (or Skipped state if Linear team has one) | When a ticket is practically blocked or out of order for today |
/tld-cancel |
Marks the current ticket Canceled and removes it from the milestone Order | When a ticket is no longer needed and should not be picked up again |
/tld-recenter |
Cuts a fresh branch off the latest default branch (detects via origin/HEAD → main → master); refuses if working tree is dirty |
After a PR merges, before starting the next ticket |
/tld-save-point |
Recovers your position from milestone + Linear | Start of a new conversation |
/tld-dashboard |
Shows progress across all milestones and tickets | When you want the big picture |
Planning
| Skill | What it does | When to use |
|---|---|---|
/campaign-plan |
Full planning — scope, phases, and tickets, all created in Linear | Starting a new project from scratch |
/milestone-create |
Creates a single Linear milestone (with optional tickets + Order) | Adding one phase without the full /campaign-plan flow |
/milestone-sync |
Authors ## Order sections on existing Linear milestones missing them |
When /tld-setup fails because Order is missing or malformed |
/tld-ticket |
Creates standardized Linear tickets (feature, bug, QA, polish) | When work needs to be tracked |
Boundaries + Side Channel
| Skill | What it does | When to use |
|---|---|---|
/tld-gate |
Full regression at milestone boundaries | When /tld-next says a milestone is complete |
/tld-side-quest |
Small fix in an isolated worktree | Polish, minor fix, or config tweak |
Campaign Config
| Skill | What it does | When to use |
|---|---|---|
/campaign-init |
Scaffolds .tld/campaign.md with the four required sections |
Setting up TLD in a new repo |
/campaign-show |
Displays the campaign's four sections plus an optional Linear snapshot | When you need to see the current project config |
/campaign-edit |
Edits a single field in .tld/campaign.md |
Updating a test command, stack path, or commit pattern |
/campaign-validate |
Schema-only check (no Linear calls) | Offline, or right after /campaign-edit to confirm the file parses |
/campaign-test |
Pre-flight connection check for the campaign (schema + Linear reachability) | Before /tld-setup; after /campaign-init |
/campaign-remove |
Deletes .tld/campaign.md (and the .tld/ dir if empty) |
Tearing down TLD config in this repo |
/campaign-portless |
Wires portless so this repo or worktree gets a stable <name>.localhost:1355 URL (installs portless / trusts certs on first run, then per-project: picks a name + free port, writes .claude/launch.json, symlinks worktree .env.local, registers the alias) |
First time on a machine, or first time in a new repo / worktree that needs its own dev URL |
Meta
| Skill | What it does | When to use |
|---|---|---|
/tld-experience |
Turn a lived conversation moment into a candidate skill — drafts a SKILL.md, updates supporting docs (CHANGELOG, tld-help, README always; STANDARDS / CONTRIBUTING / LIMITATIONS / RELEASING / ADAPTERS when applicable), opens a PR | Right after a workflow you want to be a one-liner next time |
/tld-release |
Cut a new tagged release: bumps CHANGELOG, opens a release PR, publishes the GitHub Release after merge, watches the marketplace auto-bump workflow. Optional arg: patch / minor / major / vX.Y.Z |
When you're ready to ship a new version of the plugin to /plugin update tld@claude-skills users |
Standard Flow
/tld-setup → /tld-write-tests → /tld-build ──→ /tld-audit ──→ /tld-run-test → /tld-next ──→ /tld-setup (next ticket)
| │ (optional) │ │
│ ▼ │ (fail) └──→ /tld-gate {milestoneId}
├─→ /tld-partial-auto (all-in-one, 2 gates, commits on approval) ▼ (when milestone is complete)
└─→ /tld-full-auto (hands-off → verified checkpoint) ─→ your manual check ─→ /tld-commit (per ticket) ⟶ /tld-pr (PR at story end)
/tld-align ──→ retry
Two ways to automate a code ticket. /tld-partial-auto chains write-tests → build → audit → run-test → next inside one skill with two hard stops (RED review, QA gate) and commits on approval. /tld-full-auto runs the same pipeline hands-off, stops only on real problems, and ends at a verified, uncommitted checkpoint — you do your manual check, then land it: in a multi-ticket story, /tld-commit each ticket (choose "commit and progress" to mark it Done and advance) and run /tld-pr once at the end to open a single PR for the branch; or use /tld-pr directly to commit + push + PR a standalone ticket (it stops before merge). The standalone /tld-audit step on the manual path is optional but recommended after backend / migration / auth changes.
Tips
- Type the option number (1–N depending on the block) at any "What's next?" prompt to proceed
/tld-save-pointin a new conversation replaces the old /compact paste/tld-side-questkeeps your main context clean/tld-auditis automatically marked (Recommended) by/tld-setupwhen the ticket touches the backend, has migration/auth/RLS in scope, or names an endpoint — run it then for a security pass before verify/tld-dashboardfor a quick progress check anytime/tld-helpdoesn't list itself in the reference card above (it's the card)
After printing the reference card, check the conversation context and git state to determine the user's current position, then present context-aware options:
Numbered shortcut recognition
When you present the "What's next?" options at the end of your output, the user may respond with just a number (e.g., "1" or "2"). If the user's next message is a bare number matching one of the options you presented, treat it as if they typed the corresponding slash command and invoke that skill immediately.
If there's an active ticket with uncommitted changes:
What's next?
1. /tld-commit — commit pending changes Best for: resume committing after a detour
2. /tld-dashboard — see overall progress Best for: orient yourself before deciding
3. /tld-side-quest — handle a quick fix first Best for: noticed polish worth handling before committing
Type 1, 2, or 3 to proceed.
If there's an active ticket in progress (no uncommitted changes):
What's next?
1. /tld-save-point — pick up where you left off Best for: resume the active ticket
2. /tld-dashboard — see overall progress Best for: want the big picture before diving in
3. /tld-side-quest — handle a quick fix first Best for: noticed polish worth handling before resuming
Type 1, 2, or 3 to proceed.
If no context is available:
What's next?
1. /tld-save-point — figure out where you are Best for: starting a fresh conversation, recover context
2. /tld-dashboard — see overall progress Best for: want the big picture first
3. /tld-ticket — plan a new ticket Best for: spotted work to track before diving in
Type 1, 2, or 3 to proceed.