planning-parallel-work
Skill7 starsChanged 5 months agoMITView on GitHub
Use when a substantial software task needs decomposition into dependency-aware, independently verifiable work items and safe parallel execution groups before implementation begins.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- compactSkill
- assistant-notebookSkill
- brainstormingSkill
- dispatching-parallel-agentsSkill
- executing-plansSkill
- finishing-a-development-branchSkill
- mem0-memorySkill
- next-level-agentSkill
- receiving-code-reviewSkill
- requesting-code-reviewSkill
- subagent-driven-developmentSkill
- systematic-debuggingSkill
- test-driven-developmentSkill
- using-git-worktreesSkill
- verification-before-completionSkill
- writing-acceptance-testsSkill
- writing-plansSkill
- writing-skillsSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Create an executable task graph from approved requirements and architecture.
- Distinctive: line 12
Use the available objective, requirements, acceptance criteria, architecture decisions, repository evidence, constraints, and current Git …
- Distinctive: line 26
Do not create artificial tasks for setup, documentation, or scaffolding when those changes belong to a functional deliverable.
- Distinctive: line 32
write the same files or generated artifacts.
- Distinctive: line 38
Read-only investigation may run in parallel more freely.
- Very specific: line 68
Use an empty parallel_groups array when safe concurrency cannot be demonstrated.
≈ “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
Planning Parallel Work
Create an executable task graph from approved requirements and architecture. Plan only; do not edit production code, tests, configuration, or repository state.
Inputs
Use the available objective, requirements, acceptance criteria, architecture decisions, repository evidence, constraints, and current Git state. Distinguish verified facts from assumptions. If a missing decision changes public contracts, data ownership, security boundaries, or irreversible behavior, expose it as a blocker rather than inventing it.
Decomposition
Each task must produce one independently reviewable result and include:
- a stable task ID and concise objective;
- exact acceptance criteria covered;
- files or components expected to change;
- interfaces consumed and produced;
- dependencies on other task IDs;
- verification commands and expected evidence;
- risks, unknowns, and explicit non-goals.
Do not create artificial tasks for setup, documentation, or scaffolding when those changes belong to a functional deliverable. Do not combine work that could be independently rejected by a reviewer.
Parallelism
Mark tasks parallel only when independence is supported by evidence. Parallel tasks must not:
- write the same files or generated artifacts;
- mutate the same database, service, migration sequence, lock, or external resource;
- depend on an interface or decision produced by one another;
- require ordering to preserve compatibility;
- share an exclusive test environment.
Read-only investigation may run in parallel more freely. If independence is uncertain, preserve the dependency and schedule sequentially. Never claim that a parallel group itself authorizes concurrent writes: the coordinator remains responsible for isolated workspaces, dispatch, integration order, and conflict handling.
Plan Contract
Return a concise plan followed by this machine-readable object:
{
"goal": "string",
"assumptions": [],
"blockers": [],
"tasks": [
{
"id": "T1",
"objective": "string",
"acceptance_criteria": [],
"scope": { "read": [], "write": [] },
"interfaces": { "consumes": [], "produces": [] },
"depends_on": [],
"verification": [],
"risks": [],
"non_goals": []
}
],
"parallel_groups": [["T1", "T2"]],
"integration_order": [],
"final_verification": []
}Use an empty parallel_groups array when safe concurrency cannot be demonstrated. Every acceptance criterion must map to at least one task and final verification step. Report uncovered criteria explicitly.
Handoff
The planner proposes the graph; it does not dispatch agents, approve architecture, modify files, or accept the final result. Return the plan to the coordinator for validation and execution.
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.

