agentleFS
KbediakoKbediako/evergreen-codex-skills/skills/wait-for-subagents-patiently/SKILL.md

wait-for-subagents-patiently

Skill5 starsChanged 2 months agoMITView on GitHub

Wait for quiet native Codex children and reconcile their state before declaring a stall, interrupting, replacing, or handing off without a result.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are 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.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this for native children.

    line 8
  2. Very specific:

    Ask for a checkpoint when partial evidence can change a decision.

    line 16
  3. Very specific:

    After steering, resume, compaction, or interruption, use list_agents to match children to the current scope, consume queued or final …

    line 20
  4. Distinctive:

    Read budgets and stall evidence when a deadline, resource boundary, or possible stall affects the decision.

    line 28
  5. Distinctive:

    Read interruption and ownership before a planned interruption or overlapping takeover.

    line 30

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Wait for subagents patiently

Use this for native children. External jobs use long-poll-wait; app tasks use their own event-wait tool. Load native-subagents-first for native lifecycle and run accounting unless already active. Never recursively reload it.

Wait on evidence

A wait_agent timeout means no update arrived in that window. It does not prove failure, a hang, or lack of progress. A running or completed status describes lifecycle; the actual answer and verified artifacts establish the result.

Continue independent parent work while the child runs. When its result blocks the next action, use wait_agent with the longest supported window compatible with host responsiveness and the next decision point. Do not replace mailbox waiting with sleeps or repeated lifecycle snapshots. Consume the actual mailbox update; a wake may concern another child or user input.

Ask for a checkpoint when partial evidence can change a decision. Do not request one, inspect list_agents, or narrate unchanged state solely because a timer expired. Let explicit domain patience rules govern long-running work. Never force Pro to answer early because an estimate elapsed.

Reconcile existing work

After steering, resume, compaction, or interruption, use list_agents to match children to the current scope, consume queued or final messages, and inspect expected artifacts when status is incomplete.

Use send_message for an in-scope correction to a running child. Use followup_task for authorized execution by an idle or interrupted child; messages and waits do not restart it. Count that continuation under the native-agent budget. Recover a completed answer without starting another run solely to retrieve it.

Replace only when the original is obsolete, irrecoverable, outside scope, or past a real hard decision point. Reconcile first so renamed retries do not duplicate work.

Decide whether to intervene

Read budgets and stall evidence when a deadline, resource boundary, or possible stall affects the decision. Elapsed time alone does not establish a stall. A missing required result stays an acceptance gap even when observation ends.

Read interruption and ownership before a planned interruption or overlapping takeover. Request a checkpoint in advance only when the real deadline allows a response window. Act promptly on an explicit user stop or urgent safety, permission, resource, or consequential-side-effect boundary; reconcile afterward.

Interruption does not roll back edits or cancel launched operations. Establish that prior writers and known outstanding operations have stopped or are isolated before assigning overlapping write ownership. If the interrupt result says the child had completed, recover its answer instead of calling it stalled.

Handoff

Report verified results, parent checks, missing answers, still-running work, intervention reasons, and unresolved ownership. Use status and handoff when precise lifecycle wording or a resumable handoff is needed. Do not turn status or activity into completion evidence.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.