skills.rest
alectivism
alectivismCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

subagent-delegation

Routes bounded multi-part work to the cheapest suitable subagent workers with structured return contracts.

Updated Jun 26, 2026
Search Tags:#fact-checking#codex#delegation#orchestration#subagent#chatgpt#parallel tasks
One-click install
npx skills add https://github.com/alectivism/organization-ai-skills --skill subagent-delegation-alectivism
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: subagent-delegation
Source: https://github.com/alectivism/organization-ai-skills/tree/main/org-agents-gpt/skills/subagent-delegation
Command: npx skills add https://github.com/alectivism/organization-ai-skills --skill subagent-delegation-alectivism

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and assets (resource) components.

What problem does it solve? Multi-part tasks with many searches, long documents, or repetitive items waste main-context space and cost when handled inline. This Skill enforces a delegation policy that inspects the active runtime's available subagents, routes bounded work to the cheapest suitable worker, prevents duplicate effort, and keeps synthesis and judgment in the main conversation. ## Core Features & Use Cases - Runtime-aware routing: Inspects the active tool schema before spawning anything and selects from installed agents (org-researcher, org-summarizer, org-bulk-worker, org-fact-checker, org-verifier) or built-in roles (luna-leaf, terra-ingest, sol-reviewer) without inventing unavailable models. - Structured return contracts: Defines per-worker output formats (SUMMARY/FINDINGS/UNVERIFIED, VERDICT/ISSUES, DONE/SKIPPED/SAMPLE) so delegated results arrive compact and verifiable. - Codex agent installer: Ships nine TOML agent templates under assets/codex-agents/ plus a non-overwriting bash installer for local Codex setups. - Use Case: Before drafting a report that needs five web searches and a fact-check, the main session spawns an org-researcher for gathering and an org-fact-checker for claim verification, then synthesizes the returned contracts into the final draft. ## Quick Start Apply the subagent delegation policy to plan and delegate this multi-part research task before starting it.

Frequently Asked Questions about subagent-delegation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I delegate tasks to subagents in ChatGPT or Codex?▼

Inspect the collaboration or subagent tool schema in your active session first, then assign each worker one bounded task with exclusive ownership of its sources and an explicit return contract. Only use roles, models, and effort levels the runtime actually advertises.

When should I use a subagent instead of working inline?▼

Delegate when two or more independent bounded units can run in parallel, the task needs three or more searches or reads, intermediate material would pollute the main context, or an independent read-only check reduces risk. Handle quick single lookups and voice-sensitive writing inline.

How do I install custom Codex agents from TOML templates?▼

Run the included scripts/install-codex-agents.sh, which copies the nine TOML templates from assets/codex-agents into ~/.codex/agents without overwriting existing files. Restart Codex afterward to load the newly installed agents.

Does this delegation policy work on hosted ChatGPT Work?▼

Yes, but hosted ChatGPT Work uses its own hosted agents and cannot import local TOML files. The policy falls back to whatever roles the runtime advertises and works inline when no suitable subagent capability exists.

What return format do delegated research and fact-checking workers use?▼

Researchers return SUMMARY, FINDINGS with sources and dates, and UNVERIFIED items. Fact-checkers return VERDICT, WRONG, UNVERIFIED, RIGHT BUT MISLABELED, and a checked count, and remain strictly read-only.

Related Skills

Explore More Product & Management Skills→
captainflasmrcaptainflasmr
100

blog-writer

Analyze project source, CHANGELOG, and README to generate an org-mode blog post.

Community
Advanced
kaivyykaivyy
66

perseus:report

Synthesize Perseus Scan, Audit, Exploit, and Specialist findings into a formal executive security report.

Community
Advanced
mirzadhammirzadham

santa-method

Verifies generated output through dual independent review agents with an iterative fix loop.

Community
Advanced
agentscope-aiagentscope-ai
34.3k

chat_with_agent

Connect with another agent for review or multi-turn conversation.

Official
Intermediate