tracking(codex): evaluate portable context-management ideas for PowerContext #1792

Feature description

Track portable Codex context-management ideas that PowerContext can borrow without becoming an agent runtime.

Codex prior art verified on 2026-09-29:

• [codex] Add new context window tool openai/codex#27488 - [codex] Add new context window tool
• Add history and notes tools for token-budget sessions openai/codex#39827 - Add history and notes tools for token-budget sessions
• Inject history notes hints into context windows openai/codex#40539 - Inject history notes hints into context windows
• Allow models to enable token budgeting by default openai/codex#41803 - Allow models to enable token budgeting by default
• Add experimental context management activation openai/codex#42385 - Add experimental context management activation

The portable lesson is not provider-specific session control. PowerContext should remain an agent memory and work-continuity system. The useful ideas are:

• detect when the active session is no longer a reliable carrier of work state;
• externalize current work state into evidence-backed Handoff;
• start fresh sessions from bounded, cited continuity context instead of old transcripts;
• keep transient working notes separate from durable Memory;
• evaluate whether this improves long-task continuation quality.

Problem and proposed solution

This tracking issue groups concrete PC-native follow-ups:

• feat(handoff): add host-advisory rollover triggers for long sessions #1787 - host-advisory rollover triggers for long sessions.
• feat(context): add a Handoff-backed continuation profile for prepare_context #1788 - Handoff-backed continuation profile for prepare_context.
• feat(context): inject compact continuity hints without transcript summaries #1789 - compact continuity hints without transcript summaries.
• research(work): evaluate Source-backed working notes separate from Memory #1790 - Source-backed working notes separate from Memory.
• evaluation(handoff): benchmark rollover handoffs against compaction-style summaries #1791 - benchmark Rollover Handoff against compaction-style summaries.

These issues should share a boundary:

• PowerContext does not open, close, or own model sessions.
• PowerContext does not execute a Handoff next action.
• PowerContext does not automatically promote rollover content into Memory.
• Provider-specific signals are host observations, not truth, unless backed by Source evidence.
• Existing Handoff commit/CAS/no-op semantics remain authoritative.

Alternatives considered

• Fold all Codex-inspired ideas into one implementation PR. That would mix product decisions, runtime changes, host integrations, and evaluation.
• Treat Codex's new_context as the feature to clone. That would overfit to one provider and cross PC's product boundary.
• Store session state as Memory. That would confuse temporary work state with durable project knowledge.

Additional context

Related PowerContext work:

• PR RFC: Handoff session rollover #1783 - Handoff session rollover RFC and initial compatible implementation.
• Tracking Issue: make PowerContext work continuity reliable, explainable, and cross-host #1352 - reliable, explainable, cross-host work continuity.
• tracking(continuity): 宿主历史导入与可评估的上下文连续性 #1574 - host history import and evaluable context continuity.
• feat(context): add a recall sufficiency gate with bounded expansion to prepare_context #1556 - recall sufficiency gate with bounded expansion.

Success for this tracking issue means the child issues converge on a PC-native context-continuity story: Handoff owns current work checkpoints, Source owns captured evidence, Memory owns durable knowledge, and PreparedContext owns bounded turn delivery.