Tasks

This folder is for implementation task documents.

Purpose

A task document is the AI-execution prompt for the corresponding spec — Copilot-agent ready, with file lists, sub-pass structure, verification steps, and acceptance-criteria mapping. Each task should:

• name the spec and ADRs it depends on
• restate scope and non-scope
• name touched projects or folders
• list verification steps
• cite the exact acceptance criteria it must satisfy
• contain self-contained Copilot prompts (one per pass) that paste into a fresh agent session

The canonical per-task status table lives in Roadmap Progress. This page is a flat index of every task that exists in the repo, grouped by roadmap phase.

Current state

Pre-roadmap (foundation tasks)

The four foundation slices used task packs — sets of decomposed sub-tasks for each slice. Future slices use a single TASK-*.md file with internal Copilot-pass structure rather than a pack folder.

• TASK-001: Implement First Strong Vertical Slice — completed via SLICE-001 task pack (7 sub-tasks: Foundation Shell, Domain & State Core, Simulator Boundary, Workflow Orchestration, Streaming/Diagnostics/Defects, Operator UI, Tests)
• TASK-002: Implement Persistent Run History — completed via SLICE-002 task pack (5 sub-tasks)
• TASK-003: Implement JSON Recipe File Management — completed via SLICE-003 task pack (5 sub-tasks)
• TASK-004: Implement Operational Maturity — completed via SLICE-004 task pack (6 sub-tasks)

Phase 0 — Foundations

• TASK-005: Implement CI and Quality Gates — completed (delivered via Copilot agent session)
• TASK-006: Implement Observability Baseline — completed (three Copilot passes: logging, unhandled exceptions + crash log, metrics + mutex)

Phase 1 — Simulator to scale (complete 2026-05-03)

Each Phase 1 task uses the three-Copilot-pass structure: implementation passes 1 + 2 build the slice's code; pass 3 captures the measurement row block.

• TASK-1.1: Implement Multi-Tag Telemetry — completed (3 passes)
• TASK-1.2: Implement Real Frame Payloads — completed (3 passes)
• TASK-1.3: Implement Encoder-Rate Motion — completed (3 passes)
• TASK-1.4: Implement Storm & Soak Profiles — completed (3 passes)
• TASK-1.4 Pass 3 Close-out — partial-state closeout prompt that handled the deferred Soak8h capture + final row block + runbook §4.5/§4.6
• TASK-1.5: Implement Automated Measurement Capture — superseded by TASK-1.6
• TASK-1.5.1: Implement Automated Capture Follow-Ups — superseded with TASK-1.5
• TASK-1.6: Implement FlaUI Capture — completed (3 passes; replaced the headless capture rig)

Phase 2 — Store under pressure

• TASK-2.0: Implement Store Allocation Profiling — Proposed (3 passes); Pass 1 in progress with Copilot. After it lands, the row block's Phase 2 trigger assessment determines whether 2.1 / 2.2 / 2.3 / 2.4 open.

Phase 3 — New functionality

• TASK-3.3: Implement SQLite Persistence — Proposed (3 passes); queued after TASK-2.0 lands. Three-pass shape: Pass 1 SQLite + Dapper run-history + migration runner + JSON import; Pass 2 alarm-history persistence; Pass 3 capture row + Phase 2 trigger assessment.

Conventions

• Each TASK-*.md is self-contained. The Copilot-agent prompts inside it can be pasted into a fresh agent session without external context — they reference the spec, name the files to read, and specify the acceptance criteria mapping.
• The one-pass-per-Copilot-session protocol is established by Phase 1. Each pass is a single commit. Reviewers should expect the diff to be wide on Pass 1 (most code lands), narrow on Pass 2 (extension/integration), and docs-only on Pass 3 (capture + row block + runbook).
• Index updates are batched at phase end, not per-slice. This page reflects the full Phase 1 set (complete) plus the in-progress Phase 2 / Phase 3 slices that have been opened.
• When external AI tools are used (GitHub Copilot agent mode, etc.), prefer the per-task prompts over freeform requests — they encode the project's conventions (test placement, naming, validation patterns).