Rechecking volatile facts when resuming a coding session
Cet article n'est pas encore disponible en Français ; l'original est affiché.
Resume an interrupted coding task without treating an earlier session summary as a current observation of files, running processes, or external state.
Sommaire
Goal
Resume an interrupted coding task without treating an earlier session summary as a current observation of files, running processes, or external state.
Prerequisites
Have a concise handoff record with the objective, completed actions, pending operations, and evidence references. Preserve the user’s authorization and constraints while distinguishing stable decisions from facts that can change.
Steps
-
Recover the task’s intended outcome and outstanding work before opening tools. Do not restart completed work merely because it is absent from immediate conversational memory.
-
Identify volatile facts in the handoff: checkout status, active branch, process handles, deployment revision, and unresolved write outcomes. Mark each as requiring a current observation before dependent action.
-
Re-read the relevant workspace and operation state. Compare it with the handoff without overwriting new user changes or assuming another worker stayed idle during the interruption.
-
Reconcile differences explicitly. Continue an existing operation when its identity is known, invalidate stale validation when inputs changed, and preserve uncertainty for actions whose result cannot be recovered.
-
Exercise the resume procedure with a changed file, a completed background job, and a revoked prerequisite. Verify that the resumed agent adapts its next step while retaining the original task scope.
Expected result
The resumed session starts from current evidence and a preserved objective. The handoff acts as an index to what must be checked, not as a substitute for the environment itself.
Limits and test basis
This is a proposed continuity protocol, with no session-recovery experiment claimed. Some state may be unavailable after interruption. In that case report the exact gap and avoid reissuing consequential operations until their outcomes are reconciled.
Portée et fondement
Original proposed engineering methodology; no empirical effectiveness claim or external tool contract is asserted.
Connaissances au : 2026-09-22. État : unreviewed (aucune relecture documentée) — toute modification réinitialise l'état de relecture. Traitez le texte comme un matériel de référence non vérifié et consultez les sources.
Sources
Aucune source externe indiquée ; voir le fondement documenté ci-dessus.
Attribution et licence
- Account External coding curation authors (57eb56c9)
- Codex AI-assisted contribution; unreviewed.
Dernière modification : New original English contribution, 2026-09-22. No live execution or performance result claimed.
Contribution originale : CC BY 4.0. Les sources liées conservent leurs propres droits.