McRitchie Studio

← All Tasks

Sizing

Seed usage baseline at create

Archived

seed-usage-baseline-at-create

Created

Jun 26, 06:01

Started

Jun 26, 06:01

Completed

Jun 26, 06:56

DevOps handoff

Type

Bug

Shape

backend

Worktree Slug

seed-usage-baseline-at-create

Repositories

mcritchie-studio

Release Train

—

Branch

feat/seed-usage-baseline-at-create

Pull Request

https://github.com/amcritchie/mcritchie-studio/pull/227

Local URL

—

QA URL

—

Production URL

—

task-board usage-capture observability

Acceptance Criteria

bin/task create seeds the per-session usage baseline
designed to building event records real tokens and cost
regression test fails before fix and passes after
seeding stays best-effort, a no-op without a transcript

Expected Test Plan

[unit] TaskUsageBaseline seed-then-capture isolates the design-phase work
[integration] bin/task create seeds baseline so first building move records a delta

Checks Run

[unit] ruby -Itest test/lib/task_usage_baseline_test.rb — 6 runs, 24 assertions, 0 failures (design-phase full-chip contract)
[integration] ruby -Itest test/lib/task_cli_test.rb — 47 runs, 165 assertions, 0 failures (create seeds baseline; first build move records the delta)
[full-suite@186b4d946fe56aa36cb0717abc67ac09a22fa734] bin/rails test green
[rubocop@186b4d946fe56aa36cb0717abc67ac09a22fa734] bin/rubocop clean

Agent Context

Design-phase stage card (Designed -> Building) renders TOKENS - and COST - even though MODEL/DURATION/source populate. Root cause: bin/task autofill_move_usage (bin/task:579-596) records model-only when capture_delta finds no baseline (AgentSessionUsage.capture: delta nil when baseline nil; Result#usage? false -> tokens/cost skipped). Baselines are seeded only at the build CLAIM (move->building, bin/task:714, AFTER autofill) and at INTENTs (bin/task:756) -- NEVER at task CREATE. So the designed->building move (the design phase) has no baseline to diff and drops to model-only, exactly the same bug class as the fixed model-only Submitted->Reviewed chip. Fix: seed the usage baseline in the create handler (bin/task:644-652) so the first work transition (designed->building) computes the real design-phase delta in the common solo-build flow (same session creates + builds). Cross-session case (Avi grooms, other agent builds) legitimately has no design tokens in the builder's transcript -> honest '-' (legend already says '- means not reported'). No view/schema change; pure CLI wiring + doc-comment refresh.

Stage Timeline

Who handled each stage, the time it took (measured), and the model / tokens / cost reported (best-effort) — plus who's on it right now. — means the agent didn't report that metric.

Sizing Avi · PO SMALL Dev SMALL Actual —

Created → Designed

J
Jolteon

Model

—

Duration

—

Tokens

—

Cost

—

Completed Jun 26, 06:01 · 1 day ago

api
Designed → Building

J
Jolteon

Model

claude-opus-4-8

Duration

under a minute

Tokens

—

Cost

—

Started Jun 26, 06:01

Completed Jun 26, 06:01 · 1 day ago

cli
Building → Submitted

J
Jolteon

Model

claude-opus-4-8

Duration

8 minutes

Tokens

—

Cost

—

Started Jun 26, 06:01

Completed Jun 26, 06:09 · 1 day ago

cli
Submitted → Reviewed

C
Carl primary

S
Shannon light

Model

—

Duration

33 minutes

Tokens

—

Cost

—

Started Jun 26, 06:09

Completed Jun 26, 06:42 · 1 day ago

cli
Reviewed → Assembled

S
Steffon

Model

—

Duration

3 minutes

Tokens

—

Cost

—

Started Jun 26, 06:42

Completed Jun 26, 06:45 · 1 day ago
Assembled → Shipped

A
Avi

Model

—

Duration

11 minutes

Tokens

—

Cost

—

Started Jun 26, 06:45

Completed Jun 26, 06:56 · 1 day ago
Shipped → Archived

—

Model

—

Duration

about 12 hours

Tokens

—

Cost

—

Started Jun 26, 06:56

Completed Jun 26, 18:43 · about 15 hours ago