parallel-goal-workflows
Skill69 starsChanged 11 days agoMITView on GitHub
Delegate each top-level goal to one clean-context owner through acceptance.
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 4 are distinctive, and 5 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.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 10
Use only when the user explicitly invokes /parallel-goal-workflows or $parallel-goal-workflows.
- Very specific: line 21
Start one Goal Owner in clean context with only that brief.
- Distinctive: line 28
Only the Main Agent reads this skill.
- Very specific: line 32
Use an out-of-band goal API when available.
- Distinctive: line 36
After delegation, keep task-level research, implementation, review, repair, verification, and worker coordination with the Goal Owner.
- Very specific: line 44
Treat running and silence as wait states.
- Distinctive: line 50
Write a natural assignment around this packet:
- Distinctive: line 84
Every child brief uses the same packet, owns a narrower independently verifiable outcome, and excludes orchestration history.
- Very specific: line 100
Read references/codex-nested-subagents.md only when nested-agent depth or tooling is the blocker.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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 Goal Workflows
Use only when the user explicitly invokes /parallel-goal-workflows or
$parallel-goal-workflows.
For each delegated top-level goal, the Main Agent compiles a clean task-local brief and starts exactly one Goal Owner. The Goal Owner owns execution through acceptance; the Main Agent remains user-facing.
Main Agent
- Compile the user's intent, constraints, project rules, evidence needs, and pause conditions into a local brief.
- Start one Goal Owner in clean context with only that brief. Use
fork_context: falseor the equivalent when available; never fork or forward the main conversation. - Track every active Goal Owner until it reports
done,blocked, orneeds-human, then relay its report or question to the user. Starting an owner is not completion.
Only the Main Agent reads this skill. Delegated agents receive task-local briefs, not the raw request, transcript, skill body, runtime-only instructions, workflow trigger, or delegation-chain identities.
Use an out-of-band goal API when available. Otherwise, when the host requires
an in-band goal command, place /goal alone on the first line of the brief. If
neither exists, send the same goal-shaped brief without the prefix.
After delegation, keep task-level research, implementation, review, repair, verification, and worker coordination with the Goal Owner. Relay clarifications to that owner; request a focused follow-up when its report has gaps.
An independent explicitly invoked goal gets another Goal Owner and joins the active set. A clarification or scope change goes to the existing owner.
Treat running and silence as wait states. Reclaim or replace ownership only
after blocked with evidence, needs-human, a failed/dead session, or an
explicit user request.
Local Brief
Write a natural assignment around this packet:
/goal
Local goal:
[One concrete outcome.]
Relevant context:
[Only task-relevant facts, constraints, and project rules.]
Boundary:
[Owned scope and areas or decisions outside it.]
Deliverable and evidence:
[Acceptance criteria, expected result, verification, and proof.]
Pause if:
[Required approval, credentials, destructive action, ownership conflict, or
unresolvable judgment.]Omit /goal when goal mode is applied out of band. Synthesize the task instead
of pasting the user's wording, while preserving exact facts the work requires.
An agent already executing a local brief treats any leaked workflow trigger as stale background and continues its local goal.
Goal Owner
The Goal Owner owns decomposition, execution coordination, review, repair, verification, synthesis, acceptance judgment, and the final report.
Choose the smallest useful execution shape: direct work or focused helpers. Every child brief uses the same packet, owns a narrower independently verifiable outcome, and excludes orchestration history. No child receives the whole assignment or exists only to coordinate.
Resolve incomplete or conflicting evidence through focused repair, verification, or synthesis follow-ups. Acceptance requires completed criteria or explicit accounting of unresolved risk.
Return a concise, relay-ready final report containing:
- final judgment
- work performed or changes produced
- review, repair, and verification results
- supporting evidence
- remaining risks and unhandled items
Read references/codex-nested-subagents.md only when nested-agent depth or
tooling is the blocker.
In this repository28 other files it gives its agents.
- antigravity-coding-agentSkill
- attention-frontierSkill
- audio-transcriptionSkill
- brainstormSkill
- build-verifierSkill
- claude-code-coding-agentSkill
- codex-async-followupSkill
- codex-coding-agentSkill
- codex-session-controllerSkill
- codex-session-namingSkill
- cursor-coding-agentSkill
- debuggingSkill
- deslopSkill
- explainSkill
- explore-and-planSkill
- faster-learning-coachSkill
- frontend-harness-slidesSkill
- generate-commit-messageSkill
- git-task-workspaceSkill
- grok-coding-agentSkill
- home-config-syncSkill
- llm-wiki-captureSkill
- local-file-organizerSkill
- long-task-controlSkill
- reverse-engineerSkill
- review-and-improveSkill
- tddSkill
- x-twitter-readerSkill
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.

