Ticket — variance-analysis

A mutable work-item on the rail, conforming to the brigade's TICKET-CONTRACT. Starts as the order (context pointers + scope), gets marked up as it travels the stations, ends as the full build record.

Order

Build ONE computational Finance skill encoding the standard-costing variance procedure (CMA Part 1 depth): decomposition hierarchy (static → sales-volume + flexible-budget), per-element price/rate/quantity/efficiency/spending/volume variances (DM/DL/VOH/FOH), mix & yield, and the management-by-exception interpretation step. NOT budgeting, NOT break-even. Done looks like: a trigger-tuned, procedure-first SKILL.md + progressive-disclosure references that passes the critic axes and shows measurable lift on the oracle fixtures.

Context pointers reference notes that live where the context lives (vault, 10-source-material/competencies/variance-analysis/ — in Obsidian: [[cma-variance-core-competency]], [[variance-worked-examples]], [[variance-interpretation-and-pitfalls]]). The payload carries pointers, never copies. The schema also supports url | mcp | qmd sources (see BUNDLE-SPEC); a client ticket might add a type: mcp Atlassian source for their reporting reqs.

Resolved-context snapshot

Resolved 2026-06-28 22:01 ET (vault rail backend; type: file refs → paths):

• cma-variance-core-competency.md (10-source-material/competencies/variance-analysis/) — resolved, read by spec station.
• variance-worked-examples.md (same dir) — resolved, read by test station for golden fixtures.
• variance-interpretation-and-pitfalls.md (same dir) — resolved, read for domain-fidelity/no-slop guards. (v1 records the resolution event; a production resolver stores a content hash/copy so the build replays exactly.)

Work log

• 2026-06-28 · created · queued on the vault rail.
• 2026-06-28 22:01 ET · pulled off the rail by the pass.
• 22:01 · expo phase-0 → clear · sufficient. Expo derived intent (full computational variance-decomposition skill) + type (computational) from the bundle. Non-blocking note: golden fixtures cover DM / FOH / mix-yield but not DL / VOH / sales-volume — test station to synthesize from the core competency. → proceed.
• 22:02 · spec station · read the 3 resolved sources → wrote the build spec (procedure + trigger contract + progressive-disclosure plan).
• 22:05 · test station (blind to impl) · wrote the acceptance contract from the spec + pulled golden fixtures.
• 22:09 · author station (round 1) · wrote SKILL.md + reference files.
• 22:12 · critic (advisory, 5 axes) → expo decided refire-to-author (round 1 defects to fix).
• 22:15 · author station (round 2) · revised against the expo's notes.
• 22:18 · critic (advisory) → 5/5 PASS (triggering 0.90 · domain-fidelity 0.91 · procedure-not-dump 0.92 · disclosure-hygiene · no-slop). Critics re-derived every golden fixture in Python; confirmed all 4 CMA correctness criteria.
• 22:20 · expo decided advance · rationale: bar met, every remaining nit ITERATE-grade, and on the last allowed round refiring for a cosmetic fixture gap would risk the budget wall for negligible gain. Ship.
• 22:22 ET · done · build record complete. (Skill artifacts in the run scratch; promotion to a discipline plugin pending founder review.)
• 2026-06-29 · execution-eval station (post-ship) · two-arm ablation across tiers: skill takes every tier to deterministic 100%; all lift on fixture D (judgment) — Sonnet +33pp, Haiku +22pp, Opus 0. Report: plugins/skill-agent-brigade/examples/variance-analysis/execution-eval-report.md.
• 2026-07-01 · migrated to TICKET-CONTRACT shape (context manifest → frontmatter, ## Order extracted from scope, sections canonicalized, ## Artifacts added). History preserved verbatim above; migration is itself an append event.

Artifacts

• spec / tests / critic verdicts: plugins/skill-agent-brigade/examples/variance-analysis/ (ray-plugins repo)
• the produced skill: plugins/discipline-skills/skills/variance-analysis/ (ray-plugins repo, merged in PR #6)
• execution-eval report (tier sweep, per-fixture lift): plugins/skill-agent-brigade/examples/variance-analysis/execution-eval-report.md