Track weekly progress and project risk

Where the evidence comes from

• Gold Standard: Teaching Practices | PBLWorks
• Metacognition and Self-Regulated Learning | EEF

Strong for checkpoints and coaching; AI benefit unvalidated.

1. Assemble dated work evidence

   Choose one weekly checkpoint covering at most three workstreams. Optional D05, L03 and L02 outputs supply milestones, role agreements and expectations; equivalent teacher records work. Collect short updates and inspect the work before preparing safe extracts.

   Prompt 1 · Evidence inventory Copy

   CONTEXT
   Checkpoint date, calendar version, milestone criteria and role-level agreements: [calendar]
   Minimised dated student evidence or direct teacher observations, each with work-item ID, source ID and version: [evidence]
   Teacher privacy check and freshness rule for this checkpoint, after removing names, identifying team labels, private support details and pseudonymous profiles; keep mapping offline and stop if work-item combinations identify learners: [privacy rule]
   REQUEST
   Organise the supplied work evidence against the milestones. Preserve uncertainty and source attribution.
   QUALITY BAR
   No personal or identifiable profiles. Never infer ability, wellbeing, urgency or hidden support needs. Do not grade. If dates, calendar, evidence or privacy check are missing, return HOLD and questions only. Stale evidence remains open; silence is not failure. Use no outside facts.
   FORMAT
   Use columns: work item | milestone/criterion | evidence seen | evidence date | source ID/version | gap or uncertainty. Include every supplied work item and a separate Open questions list. Missing essentials: HOLD and numbered questions only.

2. Frame questions for verification

   Compare the inventory with your originals and record corrections. The AI may group evidence gaps; you decide what they mean. Take every proposed flag to a teacher/student conversation before any action is agreed.

   Prompt 2 · Flag questions Copy

   CONTEXT
   Exact step 1 inventory checked by teacher: [approved inventory]
   Versioned teacher corrections, checkpoint date and current milestone criteria: [decisions two]
   REQUEST
   Draft one verification question for each evidenced gap. Keep existing work evidence visible.
   QUALITY BAR
   Each flag must cite dated supplied evidence or direct teacher observation; a missing-evidence flag must cite the dated record showing the gap. Unsupported or stale flags stay OPEN, not action instructions. Do not infer urgency, wellbeing, ability, motivation or personal support needs. Do not invent owners, deadlines or actions. Missing checked inventory or criteria means HOLD and questions only.
   FORMAT
   Use columns: flag ID | work item | dated evidence and source | milestone | verification question | status. Use status UNVERIFIED or OPEN. Finish with No action before teacher/student verification.

3. Record verified agreements

   Discuss every flag with students against the actual work. Record corrections, disagreements and agreed next actions outside AI. Supply only nonidentifying outcomes and role-level owners. Unsupported flags stay open without instructions; private support conversations remain offline.

   Prompt 3 · Dashboard Copy

   CONTEXT
   Exact step 2 flag table: [flags]
   Versioned teacher/student verification for every flag, corrected evidence, agreed actions, role-level owners, task-level support and review dates: [decisions three]
   REQUEST
   Build the weekly checkpoint dashboard from those supplied agreements.
   QUALITY BAR
   Never turn a flag into an action without recorded teacher/student verification and dated evidence. For any unverified or unsupported row, preserve OPEN and leave action, owner and support as not agreed. Use only supplied milestone status and project risk wording; no AI urgency, wellbeing or ability labels. No personal profiles. If the verification record is entirely missing, return HOLD and questions only.
   FORMAT
   Use columns: checkpoint date | work item | evidence seen/date/source | milestone status | next action | owner role | task support need | project risk/open question | verification record | review date. Finish with Open flags. Attribute decisions to supplied records, not AI verification.

4. Check the selected dashboard

   Select and edit the dashboard against the evidence and conversations. Audit it, then personally verify each finding before use. Keep a dated version. Stop and reduce scope if collection, conversations and review exceed 60 active minutes.

   Prompt 4 · Audit Copy

   CONTEXT
   Exact step 3 dashboard selected by teacher: [selected dashboard]
   Original dated evidence, calendar, all verification decisions and explicit final teacher edits or none: [final records]
   REQUEST
   Audit each dashboard row for evidence traceability, faithful agreements and unresolved flags. Do not rewrite or authorise actions.
   QUALITY BAR
   Require dated evidence and teacher/student verification of every flag before action. Check owners, task support and dates against supplied decisions. Unsupported or stale flags remain OPEN without action instructions. Flag personal profiles or inferred urgency, wellbeing or ability. No grading. An absent source or decision means HOLD for the affected row; do not fill gaps.
   FORMAT
   Use columns: work item | exact dashboard text | dated source/decision | finding | human check required. Cover every row. Finish with Open holds and Human verification required before action.

Check before you use it

• Can you trace each flag to dated student evidence or direct teacher observation?
• Have you verified every flag with students before agreeing any action?
• Are unsupported, stale or disputed flags open without action instructions?
• Do actions, role-level owners, support and review dates match the recorded agreements?
• Have you excluded personal profiles, identifying combinations and inferred urgency, wellbeing or ability labels?
• Have you personally checked the dated dashboard against original evidence and saved corrections?