codex-handoff
starsforks
How to install codex-handoff?
paulrberg/agent-skills/codex-handoff
$npx -y skills add PaulRBerg/agent-skills --skill codex-handoffOrchestrate one to five Codex CLI agents to implement an approved Claude Code plan.
SKILL.md
| 1 | # Codex Handoff |
| 2 | |
| 3 | Plan in Claude Code, then hand the approved implementation to one to five Codex CLI agents in the sequence the task |
| 4 | requires. The runner uses `--dangerously-bypass-approvals-and-sandbox`, so agents can write anywhere the host user can, |
| 5 | including outside the worktree. |
| 6 | |
| 7 | ## Contract |
| 8 | |
| 9 | - Run only after the user explicitly invokes this skill in Plan mode. If Plan mode is not active, ask the user to switch |
| 10 | and stop. |
| 11 | - Claude owns discovery, decisions, the implementation plan, and agent orchestration. Do not consult Codex while |
| 12 | planning. |
| 13 | - Each Codex agent implements its assigned part of the approved plan. It may inspect, edit, and validate, but must not |
| 14 | redesign the solution or return another plan. |
| 15 | - Use the smallest effective team. One agent remains valid; use additional agents only when decomposition materially |
| 16 | improves latency, correctness, or verification. Never exceed five agents total across the handoff. |
| 17 | - Keep Claude's implementation work to orchestration, integrity checks, failure handling, and the conditional polish |
| 18 | pass. |
| 19 | |
| 20 | Use `$ARGUMENTS` as the task when present; otherwise use the active user request. |
| 21 | |
| 22 | ## Plan Phase |
| 23 | |
| 24 | Produce a decision-complete plan with this section: |
| 25 | |
| 26 | ```markdown |
| 27 | ## Codex Handoff |
| 28 | |
| 29 | - Strategy: `<sequential|parallel|hybrid>` |
| 30 | - Agents: `<1-5>` — `<why this is the smallest effective count>` |
| 31 | - Validation owner: `<agent-id|claude>` — `<aggregate checks it runs once>` |
| 32 | |
| 33 | | Agent | Wave | Depends on | Scope | Model | Effort | Timeout | Implementation brief | Completion evidence | |
| 34 | | ----- | ---- | ---------- | ------------------ | ------------------------------ | ---------------------------- | ------------------- | ------------------------------------------------------ | ----------------------------------- | |
| 35 | | `A1` | `1` | `none` | `<files/behavior>` | `<gpt-5.6-terra\|gpt-5.6-sol>` | `<medium\|high\|xhigh\|max>` | `<minutes> minutes` | `<outcome, edits, constraints, and stopping criteria>` | `<commands and observable results>` | |
| 36 | |
| 37 | - Code polish: `<required|not required>` — `<reason>` |
| 38 | ``` |
| 39 | |
| 40 | Choose the execution shape from repository evidence and the approved work: |
| 41 | |
| 42 | - Use sequential agents when one agent depends on another, their write scopes overlap, or a later agent owns integration |
| 43 | or aggregate validation. |
| 44 | - Use parallel agents only for independent work with explicitly disjoint write scopes. Agents may inspect shared |
| 45 | context, but must not write outside their assigned scope. |
| 46 | - Use hybrid execution for dependency-ordered waves: run independent agents within a wave in parallel, reconcile the |
| 47 | entire wave, then start its dependents. |
| 48 | |
| 49 | A wave finishes with its slowest agent. Keep the highest-tier agent's scope minimal and move deferrable validation to |
| 50 | the validation owner. |
| 51 | |
| 52 | The five-agent limit applies to the entire handoff, not each wave. Assign every agent a stable ID, exact dependencies, |
| 53 | an implementation scope, and its own configuration and stopping criteria. If parallel work does not collectively prove |
| 54 | the overall plan, reserve a later sequential agent for integration and aggregate validation. |
| 55 | |
| 56 | Assign aggregate validation to exactly one owner per handoff: package-wide or repo-wide checks (full test suites, |
| 57 | whole-package typecheck or lint, catalog-wide checks) run once — by the integration agent when one exists, otherwise by |
| 58 | Claude during post-wave reconciliation. Every other agent's completion evidence must be the narrowest checks that prove |
| 59 | its own edits: file-scoped lint, format, or typecheck plus targeted tests for the files it touched. Duplicate aggregate |
| 60 | runs across a wave's agents are wasted wall-clock time, not extra assurance. |
| 61 | |
| 62 | Select configuration deliberately: |
| 63 | |
| 64 | | Work | Model | Effort | Baseline timeout | |
| 65 | | ---------------------------------------- | -------------------------------- | ------------------ | ---------------- | |
| 66 | | Bounded, routine implementation | `gpt-5.6-terra` | `medium` or `high` | 10 minutes | |
| 67 | | Involved multi-file implementation | `gpt-5.6-terra` or `gpt-5.6-sol` | `high` | 20 minutes | |
| 68 | | Semantic or cross-cutting implementation | `gpt-5.6-sol` | `xhigh` | 40 minutes | |
| 69 | | Exceptional, high-risk implementation | `gpt-5.6-sol` | `max` | 60 minutes | |
| 70 | |
| 71 | Never select GPT-5.6 Luna |
People who liked this also viewed
6 picksType
- find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 2026
- grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026
- vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026
- improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026
- setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026
- triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026
