发布：stock-analysis v4.16.0

CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## v4.16.0 - 2026-07-23
+
+- `--market research` 的公司与基金路径新增确定性命题发布层。每位 lens 分离 `publishable_claims` 与 `unpublished_questions`，committee 只综合达到离散证据门槛的命题；普通缺口不再被解释为看空、中性、保守或观望信号。
+- 发布命题必须引用冻结快照内的 `evidence_id`，并记录适用期、条件和失效条件；Workspace 新增 `evidence_manifest.json`、`claim_ledger.json`、`coverage_report.json` 与 `unpublished_claims.json`。
+- 价格、总市值或流动性缺失改为只阻断估值/执行行动；身份、口径完整性、一手冲突、前视偏差或完全没有可发布命题时阻断整份 research 报告。其他报告入口保持原有缺口展示契约。
+- 全面更新中英文 README 的产品定位、入口说明、架构和命题发布流程；更新双语架构 GIF、视频封面与 72 秒 Remotion 演示。
+
 ## v4.15.0 - 2026-07-20
 
 相较 v4.14.0，本版本把原有 A/HK/US 市场与 A 股公司/基金研究框架扩展为可审计的 A/HK/US/JP/KR 全球证据架构，并集中修复盘前时段、跨市场财务、基金运营、组合风险和 Agent 一手证据回填中的真实缺口。

CONTRIBUTING.md

@@ -36,6 +36,8 @@ If a public source is unavailable from your network, keep the failure visible. D
 - Prices less than or equal to zero must be filtered or marked invalid.
 - Every strong market conclusion should trace back to evidence: quote, turnover, breadth, sector rotation, flow, announcement, or risk event.
 - Source, URL, time, and fallback reason belong in evidence metadata, not hidden in prose.
+- In company/fund `research`, ordinary missing evidence filters or narrows the affected claim and is retained in audit artifacts; it must not become a bearish, neutral, conservative, or wait-and-see signal.
+- Daily, review, earnings, price-move, snapshot, and portfolio workflows keep their existing visible missing-module contracts.
 - Reports are research outputs, not investment advice.
 
 ## Pull Request Shape
@@ -69,9 +71,9 @@ Investor lenses are not roleplay prompts. A lens must define:
 
 - Evidence priority.
 - Risk vocabulary.
-- What counts as insufficient evidence.
+- What meets the discrete claim-publication rules.
 - How it changes the report structure.
-- Which data gaps must be called out.
+- Which data gaps narrow a research claim, block a claim, block an action, or block a report.
 
 Do not add a lens that mainly changes tone.
 

ROADMAP.md

@@ -1,14 +1,14 @@
 # Roadmap
 
-`stock-analysis` is already useful as a CLI. The next milestone is making it easier for strangers, AI agents, and contributors to trust it quickly.
+`stock-analysis` is available as a CLI, Python package, and generated Agent entrypoints. The next milestone is broadening primary-evidence coverage without weakening its deterministic research contracts.
 
 ## Near Term
 
-- Harden the install path: PyPI package, console script, `--help`, and first-run examples must stay in sync.
+- Keep the PyPI package, console script, `--help`, Agent entrypoints, and first-run examples in sync.
 - Keep README screenshots current with real generated output.
 - Add more compact example reports under `reports/`.
 - Improve `diagnose` output so source failures are easy to attach to issues.
-- Add agent workflow examples for daily recap automation.
+- Add compact examples for auditing the Research Workspace claim ledger.
 
 ## Data Quality
 
@@ -20,22 +20,22 @@
 ## Market Coverage
 
 - Strengthen A-share board, fund-flow, risk-calendar, and financial snapshot coverage.
-- Improve HK/US quote fallback and historical K-line consistency.
+- Continue hardening HK/US/JP/KR quote fallback and historical K-line consistency.
 - Add more fund profile fields where public routes are stable.
 - Document region-specific limitations before expanding new markets.
 
 ## Agent Workflows
 
-- Add a GitHub Actions daily recap example.
+- Keep the GitHub Actions daily recap example aligned with the current Evidence Pack contract.
 - Add a cron example for local market notes.
-- Add a Codex/Hermes/Claude Code prompt recipe that verifies conclusions against evidence JSON.
+- Expand Codex/Hermes/Claude Code recipes that verify conclusions against evidence and claim JSON.
 - Consider MCP only after the CLI and evidence contract are stable.
 
 ## Community
 
 - Keep issue templates focused on data-source failures, feature requests, lens requests, and docs/examples.
 - Submit small, well-targeted PRs to finance and AI-agent Awesome Lists.
-- Publish share images for GitHub social preview, X/LinkedIn, and Chinese technical communities.
+- Keep social previews, architecture GIFs, and bilingual demo videos aligned with each feature release.
 
 ## Out of Scope
 

agent-workflows/commands.json

@@ -10,5 +10,5 @@
   {"id":"data-diagnose","summary":"检查当前数据源、浏览器接管边界和缺失字段。","cli":"stock-analysis --market diagnose","output":"把不可用数据源、fallback 与下一步补数路径明确告知用户。"},
   {"id":"thesis-create","summary":"为一个标的创建本地、可审计的投资论文骨架。","cli":"stock-analysis --market thesis-create --symbol <symbol> --emit-evidence","output":"只写入可验证事实；假设、反证与失效条件保持为空直到用户或一手资料补充。"},
   {"id":"thesis-review","summary":"将当前 Company Evidence 与持久化论文快照比较。","cli":"stock-analysis --market thesis-review --symbol <symbol> --emit-evidence","output":"区分事实变化、证据覆盖变化和仍需人工判断的论文变化。"},
-  {"id":"research-workspace","summary":"为股票或基金生成可恢复、可审计的机构化研究工作区。","cli":"stock-analysis --market research --symbol <symbol> [--asset-type company|fund]","output":"股票冻结 C1–C8 Company Evidence，基金冻结 F1–F8 Fund Evidence；lens 与 committee 消费同一 snapshot_id，并输出证据驱动的机构报告。"}
+  {"id":"research-workspace","summary":"为股票或基金生成可恢复、可审计的机构化研究工作区。","cli":"stock-analysis --market research --symbol <symbol> [--asset-type company|fund]","output":"股票冻结 C1–C8 Company Evidence，基金冻结 F1–F8 Fund Evidence；lens 与 committee 消费同一 snapshot_id，最终报告只发布通过离散证据规则的 publishable_claims，普通缺口与未发布命题保留在四类审计 JSON。"}
 ]

assets/README.md

@@ -18,3 +18,35 @@ Screenshots and share cards used by the README and [report directory](../reports
 - `social-preview.png`: GitHub repository social preview.
 - `share-cn-1.png`: Chinese social-share card.
 - `share-x-1.png`, `share-x-2.png`: short-form X/Twitter share cards.
+
+## Product Architecture and Demo
+
+- `investor-research-architecture-en.gif`: English investor-to-report architecture animation.
+- `investor-research-architecture-zh-CN.gif`: Simplified Chinese architecture animation.
+- `diagrams/investor-research-architecture-*.source.json`: semantic inputs for the English and Simplified Chinese architecture diagrams.
+- `diagrams/investor-research-architecture-*.svg`: validated semantic SVG sources used to render the GIFs.
+- `diagrams/investor-research-architecture-*.layout.json`: geometry and layout validation reports.
+- `diagrams/investor-research-architecture-*.motion.json`: GIF motion-contract validation reports.
+- `demo-video-preview-en.png`: English poster linking to the 72-second Remotion demo.
+- `demo-video-preview-zh-CN.png`: Simplified Chinese poster linking to the 72-second Remotion demo.
+
+The architecture animations use the `fireworks-tech-graph` `agent` template and
+the `agent-orchestration` motion contract. To reproduce one locale, replace the
+placeholders below with the installed skill path and the desired locale (`en` or
+`zh-CN`):
+
+```bash
+python3 <fireworks-skill-root>/scripts/generate-from-template.py agent \
+  assets/diagrams/investor-research-architecture-<locale>.svg \
+  "$(jq -c . assets/diagrams/investor-research-architecture-<locale>.source.json)" \
+  --layout-report assets/diagrams/investor-research-architecture-<locale>.layout.json
+
+python3 <fireworks-skill-root>/scripts/fireworks.py animate \
+  assets/diagrams/investor-research-architecture-<locale>.svg \
+  assets/investor-research-architecture-<locale>.gif \
+  --preset agent-orchestration --duration 5.75 --fps 20 --width 960 \
+  --report assets/diagrams/investor-research-architecture-<locale>.motion.json
+```
+
+The generator and animator require Python 3, Node.js 18+, Chrome/Chromium,
+FFmpeg/FFprobe, `jq`, and the runtime dependencies bundled with the skill.