Documentation

Session recovery and the watchman

See how Nightshift revives a dead coding-agent session without losing the approved work contract.

Answers: recover a coding-agent session after interruption · Updated

SHORT ANSWER

The watchman observes host-specific evidence, prefers the recorded session, advances a process lease before replacement, and always hands the recovered worker the same on-disk punch list.

01

What recovery preserves

Recovery preserves the working contract first. When the host exposes a resumable identity, Nightshift continues the recorded conversation. Otherwise the files provide a controlled fresh-session handoff.

  • Approved items and gates
  • Exact next action for active walkthrough work
  • Parking-lot decisions and rollback notes
  • Snags, timestamps, and recovery events
02

Host evidence differs

Each host exposes different process and session signals, so Nightshift does not claim identical recovery mechanics. It uses only documented continuation paths and conservative fallbacks.

  • Claude Code exposes process, transcript, roster, clean-exit, pause, and API-error signals such as API Error: 500 and 529 Overloaded.
  • Codex uses process and rollout evidence and needs a resumable identity for same-conversation revival.
  • A live but unclassified wedge is left alone rather than risking duplicate work.
03

Why the old view looks stale

A headless worker can continue the session while an already-open interface cannot display appended turns. Reopen the task before interacting. The process lease prevents the stale process from performing observable tools after recovery.

TRY IT

Start with a bounded prompt

This prompt names the outcome and preserves Nightshift’s review boundary. Paste it into the supported coding host from the project you want to change.

Diagnose whether this Nightshift session is live, paused, ended, or safe to recover. Report the evidence before proposing any repair.
BOUNDARIES

What this workflow does not claim

  • Uncertain liveness evidence stands down rather than risking a second worker.
  • An already-open host interface may need to be reopened after a headless continuation.
SOURCES

Evidence and sources

These links support the released behavior, public outcomes, or problem language described on this page.

How Nightshift worksHost differences, process leases, state files, workspace layouts, guarantees, and limits.Open evidence ↗Troubleshooting decision treeRead-only diagnosis and host-specific recovery guidance.Open evidence ↗Merged Codex hardening PRAn 11-item Nightshift run in Codex that shipped in v0.9.2.Open evidence ↗
RELATED QUESTIONS

Continue from the question you have

resume Codex work after context compactionRead the answer →resume Claude Code after a crashRead the answer →diagnose a Nightshift recovery problemRead the answer →