Trackbed sync

Skill Orfi/trackbed/copilot/skills/trackbed-sync

A thin roadmap + status + orchestration layer for a Jira epic or standalone project — skills-only, for Claude Code, OpenCode, and GitHub Copilot CLI. Keep the rails, lose the train.

Install

npx -y skills add Orfi/trackbed --skill trackbed-sync

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Reconcile the Trackbed planning layer to the live roadmap — refresh the state file, roadmap, phase↔ticket mapping, and regenerate the viewer, then check every phase has a persisted plan and notify if one is missing. Read-reconcile only; never advances a phase, authors a plan, or writes Jira. Callable directly, and invoked by trackbed-orchestrate on any material change. User-invocable.

SKILL.md

3.3 KB, as published. Nobody here has run it

Trackbed Sync

Bring every Trackbed planning artifact back in step with the live roadmap in one pass. This is the single definition of the reconcile; trackbed-orchestrate calls it from its loop, and you can call it directly between turns.

<key> is auto-resolved from a single .trackbed/<key>/ directory; ask only if several exist.

Hard rules

Read-reconcile only. Refresh the planning files to match the roadmap. Never advance a phase, never author a plan, never create or link Jira. Those belong to trackbed-orchestrate, trackbed-plan, and the Jira ask respectively.
The roadmap is the single source of truth. Re-read it from disk first; if any artifact disagrees, the roadmap wins — the artifact is rewritten to match, never the reverse.
Skills-only. Read/write markdown and YAML by convention; no scripts.
Format-aware. gsd mode → .planning/STATE.md + .planning/ROADMAP.md + .trackbed/<key>/phase-jira.md; native mode → .trackbed/<key>/state.yml + roadmap.yml.

Step 1 — Read the manifest and roadmap

Read .trackbed/<key>/manifest.yml (anchor, key, format, artifact paths).
Re-read the live roadmap + state file in the recorded format.

Step 2 — Reconcile the planning layer

Bring each artifact into step with the roadmap (rewrite only what drifted):

State file — refresh current (phase + status), blockers, and session digest (stopped_at, resume_hint). Keep it lean.
Roadmap — ensure phase rows/status, owes, and insertion history match reality (gsd: .planning/ROADMAP.md; native: roadmap.yml).
Phase↔ticket mapping — keep it in sync with the roadmap (gsd: .trackbed/<key>/phase-jira.md; native: each phase's jira: field). Never write a new ticket here — that's the Jira ask in orchestration.
Viewer — regenerate .trackbed/<key>/roadmap.html by rebuilding only the DATA object from the live roadmap (delegate to / mirror trackbed-view's Step 2). The viewer is a projection, never a source.

Step 3 — Plan-presence check

For each phase that is current or todo-and-reachable, verify a persisted plan exists in the tracked planning layer (the location trackbed-plan writes to). If a phase has none, notify the user — "Phase <id> has no plan. Run /trackbed-plan <id>." Do not author the plan here; trackbed-plan owns all plan writing.

Step 4 — Report

Print a one-line summary of what changed (which files were refreshed, any plan gaps flagged), so the reconcile is auditable rather than silent.

Handoffs

Invoked directly by the user (/trackbed-sync) to reconcile on demand between orchestration turns.
Invoked by trackbed-orchestrate on any material change (a commit landing, a gate/test result, a status flip, a scope change, a blocker appearing/clearing) so its loop keeps everything current without duplicating the reconcile logic.