continue-codex-work

Resumes interrupted OpenAI Codex sessions after verifying rollout lineage and workspace state.

1.4k|216|Updated Oct 22, 2025
One-click install
npx skills add https://github.com/daymade/claude-code-skills --skill continue-codex-work
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: continue-codex-work
Source: https://github.com/daymade/claude-code-skills/tree/main/daymade-claude-code/continue-codex-work
Command: npx skills add https://github.com/daymade/claude-code-skills --skill continue-codex-work

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When an OpenAI Codex session is interrupted, compacted, forked, or stuck, picking up the work in a new agent context risks losing the original goal, repeating rejected approaches, or duplicating completed work. This Skill reconstructs the verified session lineage and rebuilds the unfinished business task before continuing.

Core Features & Use Cases

  • Verified history receipt: Requires the read-codex-history Skill to confirm the exact Session ID, fork/compaction lineage, and inherited byte boundaries before any continuation.
  • Continuation contract: Rebuilds the original business outcome, unfulfilled requests, user corrections, proven assets, and the next direct action from evidence rather than agent narration.
  • Reality reconciliation: Verifies current workspace state (branch, files, commits, external jobs) so recorded-but-failed actions are not assumed complete.
  • Use Case: A user says "Codex hit a usage limit mid-refactor, session ID abc123 — pick it up here." The Skill verifies the rollout lineage, confirms which edits actually landed, and finishes the remaining refactor without rerunning completed steps.

Quick Start

Continue the interrupted Codex session with ID abc123 and finish the task it was working on.

Frequently Asked Questions about continue-codex-work

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

FAQPage Schema
How do I continue an interrupted Codex session in a new agent context?

Provide the Codex Session ID and invoke this Skill, which first obtains a verified history receipt via read-codex-history, then rebuilds the original task contract and executes the next unfulfilled action. It never runs codex resume or codex --continue.

What is the difference between native Codex resume and this continuation workflow?

Native resume restores the same conversation's turns or compaction state directly, in which case this Skill should not be used. This Skill handles external takeover, where a different agent context must reconstruct lineage and task state from session evidence.

Can I continue a Codex session without knowing the Session ID?

Yes. The Skill delegates to read-codex-history, which offers inventory listing and bounded search to locate the correct session before producing the verified evidence briefing required for continuation.

What happens if the session lineage cannot be verified?

The Skill fails closed. Mismatched rollout IDs, missing child rollouts, broken byte boundaries, or unrecoverable parent objectives stop the continuation rather than guessing from the closest matching file.

Does this Skill trust the recorded actions from the prior Codex run?

No. Rollouts record attempted actions, so the Skill independently verifies that patches, commits, pushes, downloads, and jobs actually landed in the current workspace before treating anything as complete.