Quickstart
Overview
Thewspace skill encodes the workspace protocol for multi-repo development. It
covers worktree isolation, pipeline execution, goal-loop automation, and agent
coordination — the opinionated practices that apply to any multi-repo workspace
built on wspace.
Prerequisites and installation
Install thewspace skill into your agent workspace or global agent
environment:
Manual placement
Place theSKILL.md file directly into your workspace:
Updating the skill
What it covers
Pipeline
A single logical change from baseline to PR, with checkable exit conditions at each step:- Anchor — confirm baseline health with
wspace check - Isolate — create a worktree with
wspace worktree add - Implement — make the change inside the worktree
- Validate — run the repo’s native check suite
- Push — open a PR and watch CI to green
- Clean up — remove the worktree after merge
Goal loop
Drive many wayfinder tickets toward a destination:- FRAME — define the goal and success criteria
- SCAN — discover available tickets
- CLAIM — pick the next ticket to execute
- EXECUTE — run the Pipeline for that ticket
- REFLECT — record results, file blockers, loop or stop
Launch hierarchy
Agent startup is a race to the first useful action:- Pre-seated cwd (0 calls) — start in the worktree directly
- Handoff metadata (1 call) —
target_dirin prompt context - Root discovery (1 round trip) —
wspace check --jsonlearns everything
Token economy
- Batch commands, never per-repo probes
- Hand off plan artifacts between sessions
- Leave per-repo syntax in reference files
When to use
- Agent sessions in multi-repo workspaces
- Parallel feature development across repos
- Goal-oriented backlog execution (“clear the backlog”, “goal mode”)
- CI-driven validation workflows
How it relates to the CLI
The skill is not a thin wrapper aroundwspace. It encodes opinionated
practices — worktree isolation, pipeline steps, goal-loop structure, hard
boundaries — that apply to any multi-repo workspace. The CLI provides the
mechanics; the skill provides the protocol.