delegating-to-subagents
Skill0 starsChanged 46 days agoMITView on GitHub
Use when executing a plan whose tasks are independent enough to hand to subagents, and the harness supports dispatching them
1 other project uses a version of this.
Adopted by
Most recently taken up by
- ViktorSoroka07/claude-delivery-skillstheir own version, first seen Sep 28, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- ViktorSoroka07/claude-delivery-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository13 other files it gives its agents.
- debugging-systematicallySkill
- executing-plansSkill
- finishing-workSkill
- investigating-in-parallelSkill
- isolating-workspacesSkill
- planning-workSkill
- receiving-reviewSkill
- requesting-reviewSkill
- shaping-designsSkill
- testing-firstSkill
- using-tackleSkill
- verifying-before-claimingSkill
- writing-skillsSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
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 25
DONE: verify it (below).
- Very specific: line 26
DONE_WITH_CONCERNS: read the concerns. Resolve those about correctness before verifying.
- Distinctive: line 29
Verify independently: Read the actual diff and run the task's check yourself.
- Distinctive: line 43
Every task's diff has been read and its check run by you, review has matched each task's risk, and the full suite passes.
≈ “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
Delegating to Subagents
Hand each task to a fresh subagent with exactly the context it needs, while you act as coordinator. This keeps your own context clear for integration and judgment. Delegation costs time and tokens, so match the amount of review to the risk of each task.
Scale it
Choose per task, not per plan:
| Task risk | Implementer | Review |
| Routine (clear spec, 1–2 files) | Fast, cheaper model | You read the diff and run its check |
| Integration (several files, judgment needed) | Standard model | Diff + check, plus a reviewer if anything looks off |
| Risky (security, data, public API, hard to undo) | Most capable model | Diff + check + independent reviewer (requesting-review) |
How
- Extract every task from the plan up front so each subagent gets the full task text rather than reading the plan file.
- Dispatch one implementer at a time for tasks that touch shared files. Use
implementer-prompt.md. Include where the task fits, which interfaces it relies on, and its check. - Handle the status it reports:
- DONE: verify it (below).
- DONE_WITH_CONCERNS: read the concerns. Resolve those about correctness before verifying; note the rest.
- NEEDS_CONTEXT: supply what's missing and re-dispatch.
- BLOCKED: change something before retrying: more context, a stronger model, a smaller task, or escalate to the user if the plan is wrong. Never retry unchanged.
- Verify independently. Read the actual diff and run the task's check yourself. A subagent's report is a claim, not evidence.
- Fix proportionately. Small, clear fixes you can make directly. Larger ones go back to a subagent with the specific issue.
- Pause at checkpoints the plan marks, and whenever several tasks in a row surprise you.
- Finish with a whole-change check: full test suite, plus one review of the complete diff for substantial work.
Skip or shorten when
- The harness can't dispatch subagents (see
using-tackle/references/platforms.md). Useexecuting-plans, keeping the risk-matched checks above. - Tasks are tightly coupled or each depends on the last. Use
executing-plans. - There are only one or two small tasks. Dispatch overhead outweighs the benefit.
- The user hasn't agreed to subagent use and the harness asks you to confirm first.
Done means
Every task's diff has been read and its check run by you, review has matched each task's risk, and the full suite passes.
Next
verifying-before-claimingfinishing-work
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.