parallel-delivery
Skill0 starsChanged 2 months agoMITView on GitHub
Coordinate parallel implementation safely by isolating writers, pinning handoffs to exact commits, and preserving independent validation.
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- elixir-docsSkill
- feature-planSkill
- independent-reviewSkill
- multi-repo-changeSkill
- package-readinessSkill
- safe-refactorSkill
- security-reviewSkill
- stack-elixirSkill
- stack-nextjsSkill
- stack-phoenix-liveviewSkill
- stack-rustSkill
- stack-svelteSkill
- systematic-debuggingSkill
- technical-writingSkill
- ui-validationSkill
- unslopSkill
- worktree-integrationSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
Read-only exploration/review can share a source view when appropriate.
- Distinctive: line 20
Before dispatch, read and apply bounded coordination: one coordinator, leaf workers, finite dispatch and correction allowances, deadlines, …
- Distinctive: line 22
Start every writer from a known base SHA.
- Distinctive: line 24
Create a retained branch only for work that should survive the disposable worktree.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Parallel delivery
Parallelism is useful only when ownership is clear.
Invariants
- One writer = one worktree = one branch.
- Two writers never edit the same checkout.
- A local writer and a cloud writer never push to the same branch concurrently.
- Read-only exploration/review can share a source view when appropriate.
- Handoffs use exact commit SHAs, not "latest branch".
Procedure
Before dispatch, read and apply bounded coordination: one coordinator, leaf workers, finite dispatch and correction allowances, deadlines, verified cleanup, and explicit review completion. Keep completed run records outside Git.
- Start every writer from a known base SHA.
- Give each writer a bounded task and its own worktree.
- Create a retained branch only for work that should survive the disposable worktree.
- Require local validation before handoff.
- Record base SHA, head SHA, branch, validation, unresolved risks, and untested areas.
- Send the candidate SHA through the required independent gates.
- Integrate only reviewed SHAs.
- Discard failed worktrees/branches when no useful artifact remains.
Give each assignment an attempt identity and explicit dependencies. Track partial, blocked, ready-for-review, and reviewed states separately from whether an agent thread has finished. A delayed result or approval remains attached to its original attempt and candidate; it cannot approve a replacement.
When a writer cannot commit because Git metadata writes are denied, preserve its exact base, scoped patch including new files, and actual checks for the authorized coordinator. Require a recorded candidate commit and appropriate validation before review or integration. Do not turn the denial into an unrestricted-permission default.
A worktree prevents checkout corruption; it does not prevent semantic conflicts. If two tasks depend on an unstable shared interface, resolve that contract before maximizing parallelism.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

