Skip to content

Conversation

@bshelkhonov
Copy link

@bshelkhonov bshelkhonov commented Jan 25, 2026

Codex now supports subagents, so superpowers prompts can be updated

Motivation and Context

Previously Codex used subagents in ~20% of tasks. In other cases it printed that subagents unavailable, so I had to say it to use subagents explicitly. I've read prompts and found out that they have outdated info about subagents.

How Has This Been Tested?

I've asked Codex to create simple game and it works with subagents

Breaking Changes

No

Types of changes

  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to change)
  • Documentation update

Checklist

  • I have read the MCP Documentation
  • My code follows the repository's style guidelines
  • New and existing tests pass locally
  • I have added appropriate error handling
  • I have added or updated documentation as needed

Summary by CodeRabbit

  • Documentation
    • Task tool documentation has been updated with clear, comprehensive instructions for using subagents and executing their core operations effectively. This replaces previous messaging that indicated subagent functionality was unavailable, giving users the guidance needed to understand and leverage these capabilities in their workflows.

✏️ Tip: You can customize this high-level summary in your review settings.

@coderabbitai
Copy link

coderabbitai bot commented Jan 25, 2026

📝 Walkthrough

Walkthrough

Documentation updated to instruct usage of Codex subagents with the Task tool via spawn_agent, send_input, wait, and close_agent, replacing prior guidance that subagents were unavailable.

Changes

Cohort / File(s) Summary
Subagent Documentation Updates
.codex/superpowers-bootstrap.md, docs/README.codex.md
Replaced note that subagents are unavailable with concrete instructions to use Codex subagents via spawn_agent, send_input, wait, and close_agent.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐰 a rabbit's nibble of joy

I hopped to the docs, ears in a twitch,
Spawn, send, wait, close — now subagents switch!
Tiny codex helpers dance and play,
Guides refreshed for a brighter day. ✨

🚥 Pre-merge checks | ✅ 3
✅ Passed checks (3 passed)
Check name Status Explanation
Description Check ✅ Passed Check skipped - CodeRabbit’s high-level summary is enabled.
Title check ✅ Passed The title directly and concisely describes the main change: updating the Codex subagent mapping in the documentation/prompts files.
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✏️ Tip: You can configure your own custom pre-merge checks in the settings.


Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands and usage tips.

Copy link

@coderabbitai coderabbitai bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actionable comments posted: 1

🤖 Fix all issues with AI agents
In `@docs/README.codex.md`:
- Around line 109-111: The doc currently lists `spawn_agent`, `send_input`,
`wait`, `close_agent` and the `Skill` CLI path without clarifying they are
obra/superpowers-specific; update the README lines referencing `TodoWrite` →
`update_plan`, the Task/subagents section, and the `Skill` tool line to
explicitly state these commands and subagent flows (e.g., `spawn_agent`,
`send_input`, `wait`, `close_agent`) are provided by the obra/superpowers
extension and not part of the official OpenAI Codex API or CLI, and add a short
parenthetical note pointing users to the obra/superpowers project for details on
installing/using these utilities (keep wording minimal and adjacent to the
existing bullets).
♻️ Duplicate comments (1)
.codex/superpowers-bootstrap.md (1)

11-13: Covered in README: verify subagent tool names.

Comment on lines 109 to 111
- `TodoWrite``update_plan`
- `Task` with subagents → Tell user subagents aren't available, do work directly
- `Task` with subagents → Use Codex subagents via `spawn_agent`, `send_input`, `wait`, and `close_agent`
- `Skill` tool → `~/.codex/superpowers/.codex/superpowers-codex use-skill`
Copy link

@coderabbitai coderabbitai bot Jan 25, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🌐 Web query:

OpenAI Codex subagents spawn_agent send_input wait close_agent

💡 Result:

Those names (spawn_agent, send_input, wait, close_agent) are not part of any official OpenAI Codex API/CLI surface that OpenAI documents today.

What OpenAI does support officially:

  • Codex “subagents” = Agents SDK handoffs (delegation), not spawn_agent. In the OpenAI Agents SDK, you compose multiple agents and let them hand off work to each other (multi-agent orchestration). [1]
  • Codex CLI as an MCP server exposes only two tools:
    • codex (start a Codex session)
    • codex-reply (continue the session using threadId)
      There is no spawn_agent / send_input / wait / close_agent tool in the official Codex MCP server. [2]
  • First-class “subagent support” in the Codex repo is an open feature request, indicating it’s not currently an official capability of Codex itself. [3]

If you tell me where you’re seeing those method names (a repo link, SDK, or MCP server), I can map them to the correct official equivalents (likely: Agents SDK handoffs + codex/codex-reply).

Sources: [1] [2] [3]


🏁 Script executed:

# Check the README.codex.md file for more context around these tool names
head -n 120 docs/README.codex.md | tail -n 30

Repository: obra/superpowers

Length of output: 1020


🏁 Script executed:

# Search for any other references to these tools in the repository
rg "spawn_agent|send_input|close_agent" --type md --type txt

Repository: obra/superpowers

Length of output: 185


🏁 Script executed:

# Look for Codex configuration or tool documentation in the repo
fd -e md -e txt -e json | xargs grep -l -i "codex.*tool\|subagent" 2>/dev/null | head -20

Repository: obra/superpowers

Length of output: 878


Clarify these are obra/superpowers-specific tools, not official Codex API.

The tool names spawn_agent, send_input, wait, and close_agent are not part of the official OpenAI Codex API or CLI. These appear to be custom tools specific to the obra/superpowers framework (a third-party Codex extension). The documentation should explicitly state this to avoid misleading users into expecting these as standard Codex capabilities. Clarify that subagent support in the context of this section refers to the superpowers framework's implementation, not OpenAI's official Codex subagent/delegation features.

🤖 Prompt for AI Agents
In `@docs/README.codex.md` around lines 109 - 111, The doc currently lists
`spawn_agent`, `send_input`, `wait`, `close_agent` and the `Skill` CLI path
without clarifying they are obra/superpowers-specific; update the README lines
referencing `TodoWrite` → `update_plan`, the Task/subagents section, and the
`Skill` tool line to explicitly state these commands and subagent flows (e.g.,
`spawn_agent`, `send_input`, `wait`, `close_agent`) are provided by the
obra/superpowers extension and not part of the official OpenAI Codex API or CLI,
and add a short parenthetical note pointing users to the obra/superpowers
project for details on installing/using these utilities (keep wording minimal
and adjacent to the existing bullets).

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Fetch the content from the provided GitHub URL to verify the subagent commands
curl -s "https://raw.githubusercontent.com/openai/codex/cb2bbe5cba2c265a19d73953876bfb80f6bd44e8/codex-rs/core/templates/agents/orchestrator.md" | head -n 80

Length of output: 3804


@bshelkhonov, you're absolutely right, and I apologize for the incorrect review comment. The orchestrator template in the official OpenAI Codex repository clearly documents these subagent commands (spawn_agent, send_input, wait, close_agent) as part of the multi-agent workflow tooling.

Your PR change is correct—these are official Codex subagent capabilities. Thank you for the clarification and the direct link to the source documentation.

🐰💭
/(••)
/| |
|
___|
| |
(learning moment!)

@bshelkhonov bshelkhonov force-pushed the fix/codex-subagents-support branch from 9c293a1 to b4a59a4 Compare January 25, 2026 22:38
@deinspanjer
Copy link

Hey, looks like we both had the same idea at about the same time. :)
I have PR 361 that is very close to yours and also PR 362 that proposes some changes to the skills themselves to make use of subagents more reliable.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants