AgentStack
SKILL verified MIT Self-run

Codex Collab

skill-kevin7qi-codex-collab-codex-collab · by Kevin7Qi

Use when the user asks to invoke, delegate to, or collaborate with Codex on any task. Also use PROACTIVELY when an independent, non-Claude perspective from Codex would add value — second opinions on code, plans, architecture, or design decisions.

No reviews yet
0 installs
12 views
0.0% view→install

Install

$ agentstack add skill-kevin7qi-codex-collab-codex-collab

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Codex Collab? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

codex-collab

codex-collab is a bridge between Claude and Codex. It communicates with Codex via the codex app-server JSON-RPC protocol, giving you structured, event-driven access to Codex's capabilities — prompting, code review, tool use, and file editing.

Collaboration Modes

  • Run — Single-command run for any prompted task (research, analysis, implementation). Starts a thread, sends prompt, waits for completion, returns output.
  • Code review — Single-command review for PR-style, uncommitted, or commit-level review.
  • Parallel work — You and Codex work on different parts simultaneously. Start multiple threads.
  • Research — Spin up a read-only Codex session to investigate something while you continue other work.

Run Command (Recommended for Prompted Tasks)

The run command handles prompted tasks in one call: starts a thread, sends the prompt, waits for the turn to complete, and prints output.

# Research task
codex-collab run "what does this project do?" -s read-only --content-only

# Implementation task
codex-collab run "add input validation to the login form" --content-only

# Resume an existing thread (preserves conversation context)
codex-collab run --resume  "now check the error handling" --content-only

# Specify working directory (omit -d if already in the project dir)
codex-collab run "investigate the auth module" -d /path/to/project --content-only

IMPORTANT — Execution rules for ALL codex-collab Bash commands:

  • Always use dangerouslyDisableSandbox=true — the tool writes to ~/.codex-collab/ which is outside the sandbox allowlist.
  • For run and review: also use run_in_background=true — these take minutes. After launching, tell the user it's running and end your turn. While running: do NOT poll, block, wait, or spawn an agent to monitor — you will be notified automatically when the task finishes. If other tasks complete while Codex is running, handle them normally without checking on Codex. When notified: read the completed task's output (the notification includes it or use Read on the output file), then present the results to the user.
  • For all other commands (kill, threads, progress, output, approve, decline, clean, delete, models, health): run in the foreground — they complete in seconds.

If the user asks about progress mid-task, use TaskOutput(block=false) to read the background output stream, or:

codex-collab progress 

Code Review

For a standard PR review, call review with NO prompt string. The default pr mode runs the built-in structured diff workflow against the default branch:

# PR-style review against default branch (default — NO prompt)
codex-collab review -d /path/to/project --content-only

# Review uncommitted changes
codex-collab review --mode uncommitted -d /path/to/project --content-only

# Review a specific commit
codex-collab review --mode commit --ref abc1234 -d /path/to/project --content-only

Passing a prompt string flips to custom mode — it sends your text as free-form instructions and bypasses the built-in diff workflow. Use this when a focused or targeted review fits better than the default diff workflow (e.g., "review this for security issues", "check the error handling only"). Default to pr mode for general PR reviews:

codex-collab review "Focus on security issues in auth" -d /path/to/project --content-only

Reviews are one-shot. Each review call runs a single review inside a transient review sub-thread and exits — you cannot continue the review itself or ask the reviewer follow-up questions. For follow-ups on findings, use run --resume with the relevant review output in the prompt.

review --resume is useful for running a review with context from a task thread Codex has already been working in. It forks that context into an ephemeral read-only review thread, so the original task thread is not reconfigured or mutated. review with no --resume creates an ephemeral thread that disappears after the review — use this for standalone reviews with no prior context.

Review modes: pr (default), uncommitted, commit, custom

Context Efficiency

  • Use --content-only when reading output — prints only the result text, suppressing progress lines.
  • run and review print results on completion — no separate output call needed.
  • Use output only to re-read the full log for a previously completed thread.

Resuming Threads

When consecutive tasks relate to the same project, resume the existing thread. Codex retains the conversation history, so follow-ups like "now fix what you found" or "check the tests too" work better when Codex already has context from the previous exchange. Start a fresh thread when the task is unrelated or targets a different project.

If the user asks to continue or follow up on a prior task but you don't have the thread ID in context, follow this discovery flow:

  1. codex-collab threads --discover — see top 5 recent threads (server + local).
  2. If unsure which thread is right, codex-collab peek to see the last exchange of a candidate.
  3. For very long threads where peek alone isn't enough, spawn a subagent with codex-collab peek --limit 100 --full and ask it to summarize. This keeps the firehose out of your own context.
  4. codex-collab run --resume "..." to continue.

Only run --discover when a resume is actually wanted — it's a lookup performed on demand.

The --resume flag accepts both ID formats:

  • --resume — 8-char hex short ID (supports prefix matching, e.g., a1b2)
  • --resume — Full Codex thread ID (UUID, e.g., 019d680c-7b23-7f22-ab99-6584214a2bed)

| Situation | Action | |-----------|--------| | Same project, new prompt | codex-collab run --resume "prompt" | | Same project, want review | codex-collab review --resume | | Different project | Start new thread | | Thread stuck / errored | codex-collab kill then start new |

If you've lost track of the thread ID, use codex-collab threads to find active threads.

Checking Progress

If the user asks about a running task, use TaskOutput(block=false) (with the background task ID returned when launching the command) to read the output stream. The codex-collab thread short ID appears in the first progress line (e.g., [codex] Thread a1b2c3d4 started) — handy when you need it but don't have it. If you need just the tail of the log without the full stream:

codex-collab progress 

Note: ` is the codex-collab thread short ID (8-char hex from the output), not the Claude Code background task ID. If you don't have it, run codex-collab threads`.

Progress lines stream in real-time during execution:

[codex] Thread a1b2c3d4 started (gpt-5.4, workspace-write)
[codex] Turn started
[codex] Running: npm test
[codex] Edited: src/auth.ts (update)
[codex] Turn completed (2m 14s, 1 file changed)

Approvals

By default, Codex auto-approves all actions (--approval never). For stricter control:

# Require approval for Codex-initiated actions
codex-collab run "refactor the auth module" --approval on-request --content-only

When an approval is needed, the progress output will show:

[codex] APPROVAL NEEDED
[codex]   Command: rm -rf node_modules
[codex]   Approve: codex-collab approve 
[codex]   Decline: codex-collab decline 

Respond with approve or decline:

codex-collab approve 
codex-collab decline 

CLI Reference

Run

codex-collab run "prompt" [options]               # New thread, send prompt, wait, print output
codex-collab run --resume  "prompt" [options]  # Resume existing thread
codex-collab run "prompt" -s read-only             # Read-only sandbox

Review

codex-collab review [options]                      # PR-style (default)
codex-collab review --mode uncommitted [options]   # Uncommitted changes
codex-collab review --mode commit [options]        # Latest commit
codex-collab review --mode commit --ref      # Specific commit
codex-collab review "instructions" [options]       # Custom review
codex-collab review --resume  [options]        # Resume existing thread

Reading Output

codex-collab output                 # Full log for thread
codex-collab progress               # Recent activity (tail of log)

Thread Management

codex-collab threads                    # List threads (current session)
codex-collab threads --all              # List all threads (no display limit)
codex-collab threads --discover         # Discover threads from Codex server (top 5 by default)
codex-collab peek                   # Show last exchange (default) from server
codex-collab peek  --limit 10 --full  # Show 10 items including non-message types
codex-collab kill                   # Stop a running thread
codex-collab delete                # Archive thread, delete local files
codex-collab clean                      # Delete old logs and stale mappings

Note: jobs still works as a deprecated alias for threads.

Utility

codex-collab config                     # Show persistent defaults
codex-collab config model gpt-5.3-codex # Set default model
codex-collab config model --unset       # Unset a key (return to auto)
codex-collab config --unset             # Unset all keys (return to auto)
codex-collab models                     # List available models
codex-collab approve               # Approve a pending request
codex-collab decline               # Decline a pending request
codex-collab health                     # Check prerequisites

Options

| Flag | Description | |------|-------------| | -m, --model | Model name (default: auto — latest available) | | -r, --reasoning | Reasoning effort: low, medium, high, xhigh (default: auto — highest for model) | | -s, --sandbox | Sandbox: read-only, workspace-write, danger-full-access (default: workspace-write; review always uses read-only) | | -d, --dir | Working directory (default: cwd) | | --resume | Resume existing thread (run and review) | | --timeout | Turn timeout in seconds (default: 1200). Do not lower this — Codex tasks routinely take 5-15 minutes. Increase for large reviews or complex tasks. | | --approval | Approval policy: never, on-request, on-failure, untrusted (default: never) | | --mode | Review mode: pr, uncommitted, commit, custom | | --ref | Commit ref for --mode commit | | --all | List all threads with no display limit (threads command) | | --discover | Query Codex server for threads not in local index (threads command) | | --json | JSON output (threads, peek commands) | | --full | Include all item types in peek output (default shows messages only) | | --template | Prompt template for run command (checks ~/.codex-collab/templates/ first, then built-in) | | --content-only | Print only result text (no progress lines) | | --limit | Limit items shown |

Templates

Use --template with the run command to wrap your prompt in a structured template.

Custom templates: place .md files with frontmatter in ~/.codex-collab/templates/, then re-run the installer.

TUI Handoff

To hand off a thread to the Codex TUI, look up the full thread ID with codex-collab threads --json and then run codex resume in the terminal.

Tips

  • run --resume requires a prompt. review --resume works without one (it uses the review workflow), but run --resume will error if no prompt is given.
  • Omit -d if already in the project directory — it defaults to cwd. Only pass -d when the target project differs from your current directory.
  • Multiple concurrent threads are supported. Threads share a per-workspace broker for efficient resource usage.
  • Validate Codex's findings. After reading Codex's review or analysis output, verify each finding against the actual source code before presenting to the user. Drop false positives, note which findings you verified.
  • Per-workspace scoping. Threads and state are scoped per workspace (git repo root). Different repos have independent thread lists.
  • First invocation per workspace may take slightly longer to initialize; subsequent calls in the same session reuse the connection context.

Error Recovery

| Symptom | Fix | |---------|-----| | "codex CLI not found" | Install: npm install -g @openai/codex | | Turn timed out | Increase --timeout (e.g., --timeout 1800 for 30 min). Large reviews and complex tasks often need more than the 20-min default. | | Thread not found | Use codex-collab threads to list active threads | | Process crashed mid-task | Resume with --resume — thread state is persisted | | Approval request hanging | Run codex-collab approve or codex-collab decline |

Prerequisites

Requires bun and codex CLI on PATH. Run codex-collab health to verify.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.