agentsclimarketplace

Parallel worktree supervisor

Skill ldilov/harness-forge/skills/parallel-worktree-supervisor

AI coding agent runtime that makes LLMs understand your codebase with repository-aware context, memory, and continuous learning

Install
npx -y skills add ldilov/harness-forge --skill parallel-worktree-supervisor

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

plan and supervise low-overlap parallel work across git worktrees or stacked branches. use when a backlog could be split safely, when multiple small dependent changes need reviewable sequencing, or when merge-blocking risks must be made explicit before parallel execution.

SKILL.md

2.6 KB, as published. Nobody here has run it

Parallel Worktree Supervisor

Trigger Signals

  • the backlog could be split into low-overlap tracks with clear ownership boundaries
  • multiple small dependent changes need stacked review instead of one giant branch
  • the team wants explicit merge blockers, rebase rules, and validation gates before parallel execution

Inspect First

  • task list or issue graph, including dependencies and shared prerequisites
  • repo hotspots such as root config, lockfiles, generated artifacts, migrations, and shared contracts
  • code ownership, test entrypoints, and branch protection or merge expectations
  • whether worktrees, stacked branches, or plain feature branches are already in use

Workflow

  1. map tasks by dependency, risk, and shared-file overlap
  2. carve the work into tracks that minimize contention on root config, generated files, and public contracts
  3. assign a branch or worktree plan with base branch, stack order, and merge blockers
  4. define rebase cadence, validation gates, and when a track must pause for another track to land
  5. report status in terms of ready, blocked, risky overlap, or merge-ready

Output Contract

  • sharding plan with branch or worktree names and ownership notes
  • explicit dependency graph and merge order
  • overlap hotspots and the files or contracts that make them risky
  • validation and rebase expectations for each track

Failure Modes

  • hidden coupling means the work is not actually independent enough to parallelize
  • the repo relies on shared generated files, lockfiles, or root config that every track must touch
  • there is no trustworthy validation path to catch skew between branches

Escalation

  • escalate when a schema, API, or root-build change affects multiple tracks at once
  • escalate when stacked branches become too interdependent to review independently
  • escalate when branch protection, review ownership, or release timing constraints are unclear

References

  • skills/parallel-worktree-supervisor/references/task-sharding-rules.md
  • skills/parallel-worktree-supervisor/references/worktree-operations.md
  • skills/parallel-worktree-supervisor/references/stacked-diffs-and-review.md
  • skills/parallel-worktree-supervisor/references/merge-readiness.md
  • skills/parallel-worktree-supervisor/references/output-template.md
  • skills/parallel-worktree-supervisor/references/examples.md

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.