Scenario book: multi-day scheduled research sprint¶
Goal¶
Write a scenario showing a researcher using worklog scheduling to run a systematic milestone-clearing sprint over multiple days, with automated orient → execute → record cycles and human checkpoints between days.
Context¶
This scenario exercises the worklog + questio scheduling workflow — the most autonomous pattern, where background agents run research sessions on a schedule and the researcher reviews progress daily. It demonstrates how the orient loop, iterate loop, and worklog scheduling compose for sustained autonomous research.
The scenario should be written at docs/specs/research-orchestration/scenarios/scenario-research-sprint.md.
Pixecog context:
- After TTL removal is validated, multiple milestones are unblocked: ieeg-preprocessing-stable, ecephys-preprocessing-stable, brainstate-classification-validated
- Each is relatively straightforward (configure flow, run pipeline, validate outputs, record evidence)
- The researcher wants to clear all three preprocessing milestones over a week while focusing on other work
- Worklog tools: schedule_queue, enqueue_task, list_queue, agenda, focus
- The researcher reviews progress each morning via questio_status and questio-report
- Background agents follow the iterate loop pattern but with propose-review-confirm — they create result notes and PROPOSE milestone updates, they don't auto-apply
Prompt¶
Write the scenario at /storage2/arash/projects/projio/docs/specs/research-orchestration/scenarios/scenario-research-sprint.md.
Step 1: Read context.
- Read /storage2/arash/projects/projio/docs/specs/research-orchestration/scenario-book.md for format
- Read /storage2/arash/projects/projio/docs/specs/research-orchestration/loop-mechanisms.md
- Read pixecog milestones.yml for the preprocessing milestone chain
Step 2: Write the scenario.
Structure:
1. Header with starting state: ttl-removal-validated is complete, three preprocessing milestones are unblocked, researcher wants to clear them over a week
2. Day 1 morning: Planning session — researcher says "I want to clear all preprocessing milestones this week. Can you set up a sprint?" Agent orients (questio_status, questio_gap), identifies the three targets, checks pipeio_flow_status for each. Proposes a schedule:
- Day 1: iEEG preprocessing (remaining steps after TTL)
- Day 2: ecephys preprocessing
- Day 3: brainstate classification
- Day 4: review all results, update milestones
3. Day 1: Scheduling — Agent creates task notes for each day's work using worklog_note(kind="issue") with detailed prompts. Schedules them via schedule_queue with time-based triggers (each morning at 03:00) or dependency-based (after previous task). Shows the actual schedule_queue calls.
4. Day 1 evening: Background agent runs — The scheduled agent runs the iEEG preprocessing pipeline: orients → grounds → runs pipeline → assesses → creates observation notes → creates result note → PROPOSES milestone update (does not auto-apply). Agent sends notification.
5. Day 2 morning: Review — Researcher opens a new session. Agent calls questio_status and list_queue. Shows: "iEEG preprocessing completed overnight. Result note created. Milestone update proposed — ieeg-preprocessing-stable → complete. Pending your review." Researcher reviews the result note, approves the milestone update. Ecephys task is already running.
6. Day 3 morning: Course correction — Ecephys preprocessing had an issue overnight (one subject failed). Agent reports the failure and observation notes from the background agent's investigation. Researcher says "skip sub-04 for now, it has known electrode issues." Agent updates the brainstate task to exclude sub-04 and re-schedules.
7. Day 4: Sprint review — All three milestones addressed. Agent runs questio-report for the week. Shows: milestones completed, milestones with caveats (sub-04 excluded), what's unblocked next (event detection milestones). Researcher decides on next sprint priorities.
Use mkdocs material admonitions:
- !!! info "Behind the scenes" for tool calls and scheduling details
- !!! tip "Autonomy levels" for explaining what background agents can and cannot do
- !!! warning "Human checkpoint" for morning reviews and milestone approvals
- !!! danger "Course correction" for the Day 3 failure scenario
- !!! note "Scheduling pattern" for explaining schedule_queue with after= dependencies
- !!! example "Morning report" for questio-report output
End with: ecosystem coverage (worklog + questio + pipeio — the scheduling triad), loop patterns (orient → scheduled iterate → human review cycle), key insight (answer: autonomous research sprints work when background agents PROPOSE and humans APPROVE — the daily review is the essential human-in-the-loop checkpoint that prevents autonomous drift).
Step 3: Commit with message: "Add scenario: multi-day scheduled research sprint"
Acceptance Criteria¶
- [ ] File at
docs/specs/research-orchestration/scenarios/scenario-research-sprint.md - [ ] Shows worklog schedule_queue with time and dependency triggers
- [ ] Background agents follow propose-review-confirm
- [ ] Includes a failure/course-correction day
- [ ] Morning review pattern is explicit
- [ ] Uses mkdocs material admonitions
- [ ] Committed
Batch Result¶
- status: done
- batch queue_id:
140acc720c2b - session:
49cfe4a4-561d-4bb8-8fb6-d01b2ac17020 - batch duration: 801.3s