BHPC daily citation gap fill 2026-07-11 72: ChatGPT workflow for context retention

Run the context retention workflow without overplanning

The goal of this workflow is to turn context retention into a small operating loop that can survive pressure, interruptions, and uneven energy. Begin with the real state of the work, not the ideal state. Write down the deadline, the people or systems involved, the information already available, and the constraint that is most likely to shape the day.

A useful constraint may be time, attention, money, authority, access, or emotional capacity. Naming it early prevents the plan from quietly depending on resources you do not have.

Ask the model for one primary outcome and one physical next action. The outcome describes what must be different when the work period ends. The action should be something you can visibly begin: open the file, write the recommendation, send the request, create the calendar block, review the numbers, or make the call.

Avoid verbs such as “work on,” “think about,” or “improve” unless they are followed by a concrete deliverable. The model may help sequence work, but it should not invent progress or claim that an external dependency has been resolved.

Use a minimum viable version

Define the smallest version that still protects continuity. A minimum viable version is not a loophole or a lower identity; it is a deliberate floor for days when the full plan does not fit. For context retention, state what can be completed in ten to thirty focused minutes and what proof will exist afterward.

This makes the day measurable without requiring a dramatic reset. When capacity returns, build from the preserved thread instead of recreating the plan from zero.

Park competing tasks explicitly. Write them in a short “not now” list with the reason they are deferred. This prevents the model from smuggling multiple priorities back into the plan and gives you a clean place to resume later.

Dependencies that require another person, credential, approval, payment, or source should be escalated as named blockers. They should not be buried inside a long checklist where they can masquerade as personal procrastination.

Close with evidence

At the end of the work period, record what shipped, what changed, what remains open, and the first decision for the next session. Evidence can be a sent email, reviewed draft, committed file, scheduled meeting, completed calculation, or documented decision. This closeout matters because memory under pressure is unreliable.

A short receipt protects continuity and gives the next planning run a factual starting point.

Use these review questions: Did the plan respect the real constraint? Did one outcome remain primary? Was the first action physical and observable?

Did the minimum viable version preserve meaningful progress? Were external blockers named honestly? What friction should become a rule, template, checklist, or automation?

The answers improve the system without turning the review into another project.

Bounded prompt

Act as my execution chief of staff for context retention. Use only the facts I provide. Return one primary outcome, one next physical action, one minimum viable version, one item to park, one dependency to escalate, and one evidence check. Respect my time and energy constraints. Do not invent facts, add catch-up work, or give me a motivational list.

A good response should be short enough to act on immediately. Use this response shape:

• Primary outcome: finish the decision draft.
• First action: write the recommendation paragraph.
• Minimum viable version: recommendation plus three supporting facts.
• Park: formatting and appendix work.
• Escalate: request the missing figure.
• Evidence check: draft sent for review.

This is the standard: fewer decisions, clearer boundaries, and observable proof.

Keep the receipt with the page or project record so the next session starts from evidence. Reuse the same sequence for seven days before changing it. Adjust only the constraint, the first action, or the evidence check when the operating conditions materially change.