| name | parallel-lane-orchestrator |
| description | Use when multiple dependency-ready tickets can advance simultaneously and we need explicit execution waves, lane ownership, and serialized-resource rules before work starts. |
Parallel Lane Orchestrator
Use this skill when the question is:
- which tickets can run at the same time
- where are the true parallel lanes in today's DAG
- what resources force serialization
- how do we avoid review chaos later
This skill plans multi-ticket execution waves.
It should compose with worktree-lane-orchestrator, not replace it.
Inputs to gather first
- selected tickets for today
- dependency graph
- known shared blockers
- file-surface overlap if already known
- runtime collisions and serialized resources
- expected review convergence points
Start by reading:
../worktree-lane-orchestrator/SKILL.md
../ticket-to-human-review/SKILL.md
../review-batch-orchestrator/SKILL.md
../references/git-merge-close-gate.md
Goal
Create one honest lane map for today's tickets:
- truly parallel tickets grouped together
- serialized tickets called out explicitly
- shared blockers and shared resources visible up front
- review convergence planned before execution begins
For DB-proof campaigns, the lane map must also keep frozen phase order stronger than "dependency-free
and looks easy." Do not let broad tagged sets blow up the slice.
Workflow
1. Partition tickets into waves
Group tickets by dependency readiness:
- Wave A: no unresolved blockers
- Wave B: unlocks after Wave A
- later waves only when justified by today's capacity
For DB-proof campaigns:
- wave by frozen phase and bundle first, then by ticket-level blocker edges
- start with the proof-substrate and receipt-truth foundation squads
- then repair Bundle A before flowing into the rest of Phase 1
- only after that should later bundles enter active execution waves
2. Partition each wave into lanes
For each wave:
- group tickets that can run independently
- split tickets that would collide on the same critical surface
- mark any runtime-verification lane that must serialize
Typical serialized resources:
ui_automation_driver verification
- one shared staging deploy surface
- one integration branch that cannot absorb multiple competing mutations safely
- one shared DB reset /
db_runtime state surface when persisted-path proof would otherwise collide
3. Emit lane ownership
Every lane should include:
- owned tickets
- owner
- expected finish condition
- serialized resources
- review handoff condition
- merge-to-main handoff condition after human
VERIFIED
For DB-proof campaigns, default lane families are:
- campaign director / trust-table lane
- proof-substrate lane
- receipt-truth lane
- ticket execution lanes by owning surface
4. Prepare review convergence
For each wave, declare:
- whether tickets will go to one grouped review wave
- whether any ticket must be reviewed alone
- which patch loop follows if review findings land
Close rule
This skill is complete only when:
- today's tickets are grouped into honest parallel waves
- serialized resources are explicit
- lane ownership is explicit
- the execution map is clean enough for
worktree-lane-orchestrator and review-batch-orchestrator to pick up next
- any ticket that will eventually close has an explicit path from review to merge-to-main instead of
assuming human signoff equals
Done