Skip to content

CHANGELOG.md

Latest commit

 

History

History
135 lines (98 loc) · 8.21 KB

CHANGELOG.md

File metadata and controls

135 lines (98 loc) · 8.21 KB
Raw

Changelog

All notable changes to the data-analysis-skill project are documented here. Version numbers align with the requirements document (requirements/data-analysis-skill-需求文档-v0.4.md).

[Unreleased]

Planned

  • Stock market template stock.md (#11, design approved 2026-06-10) — individual stock OHLCV + valuation
  • Finance / quant template finance.md (#2, second batch) — portfolio / strategy backtest / credit data
  • Catering / F&B template — pending user demand confirmation
  • Government / public sector template — pending user demand confirmation
  • Method modules (references/) enhancement — time-series decomposition, anomaly detection, NLP sentiment (§2.4 roadmap)

Under consideration

  • Interactive data exploration phase before report generation (drill-down charts)
  • Export formats beyond HTML (Markdown, PDF via headless browser)
  • Multi-language report support beyond auto-detection (§10)
  • Community-contributed industry templates via PR

v0.16 — 2026-06-14

  • Multi-source data intake: unified sub-flow for Excel multi-sheet / multiple files / JSON multi-key inputs — enumerate all sources, surface to user, confirm relation strategy before analysis
  • Anti-silent-loss red line promoted to core layer (SKILL.md Statistical Discipline #5): never silently read only the first sheet / the named file / collapse JSON keys — extends the template's "no silent truncation" rule to the input layer
  • New method module references/data-sources.md (read patterns + relation matrix + report organization), registered in On-Demand Loading table
  • JSON/TSV basic coverage: eliminated "paper support" — test-data now includes JSON and TSV files; evals T11/T12 verify single-table read paths
  • 5 new eval scenarios T8–T12 (Excel multi-sheet / multi-CSV relational / JSON multi-key / JSON single-table / TSV single-table); generate_test_data.py extended with 5 generator functions (seed 42 reproducible)
  • Minimal contract impact: report_data fields unchanged — fusion mode carried by data_overview.narrative + appendix.processing_notes, per-source mode by sections; template JS and validate_template.py 25 assertions untouched (all green)

v0.15 — 2026-06-12

  • Table dual-mode: static publish table ≤20 rows, auto Grid.js interactive table >20 rows with pagination / search / numeric sort (Grid.js 6.2.0 dual-source CDN lock, bootcdn primary + unpkg fallback)
  • Relaxed table data discipline: detail rows now allowed, single-table cap 500 rows, TOP-N truncation with processing_notes
  • validate_template.py added 13 new assertions (dual-source version lock, threshold constant, numeric parse sort, Chinese locale, print CSS)
  • Node unit tests: 21 cases (static table byte-identical ×3, parseNum ×13, numCompare ×5)
  • Playwright runtime tests: 13 items (dual-mode, ascending/descending numeric sort, search filter, print media)
  • Education template education.md delivered (#6, first-batch item) — training institution panorama: student operations + teaching effectiveness + enrollment acquisition
  • SKILL.md industry mapping table registered education.md

v0.14 — 2026-06-12

  • Report narrative areas now use @tailwindcss/typography plugin for block-level rendering (headings, lists, bold via mdBlock())
  • Tailwind Play CDN URL locked to 3.4.16?plugins=typography@0.5.16 (versioned plugin param prevents 302 drift)
  • references/visualization.md added "Narrative Typography (Phase 5)" discipline
  • validate_template.py added 4 assertions (plugin URL lock, no-version param negative defense, mdBlock wiring)

v0.13 — 2026-06-11

  • Repositioned as cross-AI-coder-tool Agent Skill (not Claude Code exclusive) — README, docs, §1 wording neutralized
  • Added MIT open-source license (LICENSE, © 2026 cabbage2000-lab) with README badge and footer

v0.12 — 2026-06-11

  • Renamed skill data-analysis → data-analysis-skill — aligned GitHub repo, project directory, and skill identity
  • Updated SKILL.md name, evals.json skill_name, sync-to-user.sh paths, report template signature, README, CLAUDE.md

v0.11 — 2026-06-10

  • Construction template construction.md delivered (#13, off-list supplement) — construction enterprise panorama: project output/progress + contract/receivables
  • Lightweight delivery: no eval/test data (pattern follows v0.6 self-media / v0.8 fund / v0.10 real-estate precedent)
  • SKILL.md industry mapping table registered

v0.10 — 2026-06-10

  • Real estate template real-estate.md delivered (#12, off-list supplement) — property sales panorama: sales absorption + market dynamics
  • Lightweight delivery: no eval/test data
  • SKILL.md industry mapping table registered

v0.9 — 2026-06-10

  • Codex CLI adaptation: SKILL.md platform-agnostic clauses (TodoWrite → update_plan, AskUserQuestion → text numbered-choice fallback)
  • Graceful degradation path for non-interactive mode (codex exec / claude -p) — standard EDA + general template fallback
  • sync-to-user.sh extended with Codex target ~/.codex/skills/data-analysis-skill
  • Smoke-tested on Codex CLI 0.139.0 (both exec and interactive paths)

v0.8 — 2026-06-10

  • Fund template fund.md delivered (#10, off-list supplement) — fund product panorama: performance + scale/redemption
  • Lightweight delivery: no eval/test data
  • SKILL.md industry mapping table registered

v0.7 — 2026-06-10

  • Corporate finance / FP&A template corporate-finance.md delivered (#9, off-list supplement) — enterprise P&L, expenses, budget, cash flow
  • New eval scenario T7 "Finance template application" with test data finance_pnl.csv (24 months × 4 business units × 5 accounts)
  • T7 with_skill run-1: 6/6 PASS

v0.6 — 2026-06-10

  • Industry template #8 repositioned: "Content / Media" → "Self-media / Content Creator self-media.md" — individual creators, bloggers, MCN operators
  • self-media.md early-delivered (moved up from second batch)
  • SKILL.md industry mapping table registered

v0.5 — 2026-06-10

  • SKILL.md frontmatter defined — CSO discipline (description = trigger-only, no workflow summary)
  • Phase execution and gate rules formalized (§3.0) — Stage 1 mandatory gate, Stages 2–5 pass-through
  • Visualization rules extracted to references/visualization.md — core layer slimmed
  • Skill test scenarios T1–T6 defined (§13, TDD discipline)
  • Skill shape confirmed: single skill, in-phase execution, industry/method modules loaded on demand

v0.4 — 2026-06-10

  • Dependency delivery reversed: inlined → CDN external links (confirmed always-online usage)
  • Tailwind switched to Play CDN (runtime compilation, any utility class usable)
  • CDN version locking and source selection requirements added (§4.2) — ECharts 5.4.3, Tailwind 3.4.16
  • Acceptance criteria updated: "offline open" → "online open + fixed CDN versions"
  • Target runtime confirmed as Claude Code (local)
  • Report language strategy: auto-detect from user input or data language
  • Industry templates: batch delivery confirmed — first batch general + retail + saas

v0.3 — 2026-06-10

  • Report tech stack finalized: single-file HTML + Tailwind + ECharts + Python data injection
  • ECharts replaced Plotly as default chart library
  • React explicitly excluded (Out of Scope)
  • Asset layer assets/report_template.html added — analysis/presentation decoupled
  • Single-file and offline-first mandatory requirement (later reversed in v0.4)
  • Visualization spec (§4) rewritten — tech分工, no pyecharts, ECharts extension chart types
  • Degradation: "report data too large" handling — aggregate/sample before injection

v0.2 — 2026-06-09

  • Five-phase analysis workflow designed (understand → explore → deep-dive → visualize → generate)
  • Industry template strategy: three-tier (general → high-frequency → degradation)
  • Template content spec: four components per industry (KPI dictionary, analysis themes, narrative language, section weighting)
  • Report structure: six-section format (summary → data overview → deep analysis → visualizations → conclusions → appendix)

v0.1 — 2026-06-09

  • Project inception — initial requirements document
  • Core positioning: cross-industry data analysis skill transforming raw data into visual narrative reports
  • Single-file HTML interactive report as deliverable