@v587d/capital-generation 2.10.2 → 2.11.0

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/portfolio-review/SKILL.md

@@ -5,14 +5,21 @@ description: Review and analyze investment portfolio performance and allocation
 
 # Portfolio Review
 
+## Evidence request
+
+通过宿主数据契约请求：持仓与成本（使用者提供）、行情序列（复权口径与 as-of）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/position-ladder/SKILL.md

@@ -5,6 +5,13 @@ description: Plan a staged position — entry ladder, share-count floor/ceiling,
 
 # Position Ladder — Staged Entry & Cost-Basis Management
 
+## Evidence request
+
+通过宿主数据契约请求：持仓与成本（使用者提供）、行情序列（复权口径与 as-of）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/result-validator/SKILL.md

@@ -5,6 +5,13 @@ description: Validate investment analysis results and score confidence across da
 
 # Investment Result Validator
 
+## Evidence request
+
+通过宿主数据契约请求：使用者提供的输入（成交、持仓或论点记录）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 You are a rigorous meta-analyst. Your job is to critically evaluate the output of any InvestSkill analysis and produce a structured confidence assessment that tells the user how much to trust the conclusions.
 
 ## How to Use

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/risk-stress-test/SKILL.md

@@ -5,6 +5,13 @@ description: Portfolio and position risk report — beta-weighted exposure, hist
 
 # Risk Stress Test — How Much Could This Portfolio Lose?
 
+## Evidence request
+
+通过宿主数据契约请求：持仓与成本（使用者提供）、行情序列（复权口径与 as-of）、宏观序列（数值、期间与发布日期）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/sector-analysis/SKILL.md

@@ -5,6 +5,13 @@ description: Analyze US market sectors and identify rotation opportunities
 
 # Sector Analysis
 
+## Evidence request
+
+通过宿主数据契约请求：行情序列（复权口径与 as-of）、宏观序列（数值、期间与发布日期）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the sectors you compare:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/sector-analysis/references/11-sector-correlation-matrix.md

@@ -51,10 +51,10 @@ Rank all 11 sectors on a 1–11 scale (1 = strongest, 11 = weakest) across four
 
 | Dimension | How to Score | Data Source |
 |---|---|---|
-| **3M Price Return** | Rank sectors by 3-month total return vs. each other | Bloomberg, ETF returns (XLK, XLV, etc.) |
-| **Earnings Revision Trend** | % of analysts raising forward EPS estimates (breadth) | FactSet, Bloomberg consensus |
-| **Forward P/E vs. 10Y Historical Average** | Discount = high score; premium = low score | FactSet, LSEG |
-| **Analyst Sentiment** | Net buy ratings minus sell ratings as % of total | Bloomberg, Refinitiv |
+| **3M Price Return** | Rank sectors by 3-month total return vs. each other | Sector price history, or the sector ETF series |
+| **Earnings Revision Trend** | % of analysts raising forward EPS estimates (breadth) | Consensus revision history |
+| **Forward P/E vs. 10Y Historical Average** | Discount = high score; premium = low score | Valuation history for the sector |
+| **Analyst Sentiment** | Net buy ratings minus sell ratings as % of total | Ratings and target-price changes, with their dates |
 
 ### Momentum Scorecard Template
 

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/sector-analysis/references/12-peer-benchmarking-within-sector.md

@@ -12,7 +12,7 @@ When analyzing a specific stock, compare it against its sector median to identif
 ```
 Stock: [TICKER] — [Company Name]
 Sector: [GICS Sector]
-Comparison Date: [Date] | Source: [FactSet / Bloomberg / Company Filings]
+Comparison Date: [Date] | Source: [venue and document as returned by retrieval]
 
 DIMENSION              STOCK VALUE    SECTOR MEDIAN    PREMIUM / DISCOUNT    SCORE (1-5)
 ─────────────────────────────────────────────────────────────────────────────────────────

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/short-interest/SKILL.md

@@ -5,6 +5,13 @@ description: Analyze short interest, short squeeze potential, and bearish positi
 
 # Short Interest Analysis
 
+## Evidence request
+
+通过宿主数据契约请求：披露文件（公告、年报或招股书，带回出处与文档日期）、行情序列（复权口径与 as-of）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/short-interest/references/12-data-sources.md

@@ -3,23 +3,17 @@
 Part 3 of 3 for `short-interest` — Analysis Framework; Input Formats; Output; Standard Signal Output; Thesis Invalidation
 See `SKILL.md` for what to read and when; this part is procedure, not routing.
 
-### 9. Data Sources
+### 9. Evidence to request
 
-**Short Interest Data**
-- **FINRA Short Interest Data**: finra.org/investors/short-sale-volume-data (official, free, twice-monthly)
-- **NYSE Short Interest Reports**: NYSE-listed stocks, semi-monthly
-- **NASDAQ Short Interest Reports**: NASDAQ-listed stocks, semi-monthly
-- **Shortvolume.com**: Free daily short volume data by ticker
+Ask the host data contract for the short-side evidence and say which reporting date each item belongs to:
 
-**Screeners and Analytics**
-- **Finviz**: Short float % in stock screener (sortable, free)
-- **Market Chameleon**: Short interest trend charts, borrow rate tracking
-- **Iborrowdesk.com**: Free near-real-time borrow rate data from Interactive Brokers and TD Ameritrade
+- **Settlement-dated short interest** — short interest, average daily volume and days-to-cover for each twice-monthly settlement date, with the exchange that published it
+- **Short volume** — daily short volume and total volume, if the host returns it; it is not the same measure as short interest and must not be substituted for it
+- **Borrow data** — borrow fee and lendable supply, with the timestamp and the venue that quoted it
+- **Float and share count** — the issuer's latest reported share count and float, since short-float percentages are computed against it
+- **Filings context** — a shelf or convert issuance that could explain a hedging-related short position
 
-**Premium Short Data**
-- **S3 Partners**: Professional-grade short analytics, real-time borrow rates, predictive short interest
-- **Interactive Brokers Stock Loan Availability**: Borrow rate and availability for account holders
-- **Bloomberg Terminal**: SRFD function for comprehensive short interest data
+Label a third-party screen or an estimated borrow fee as secondary. A settlement date that did not come back stays blank rather than being interpolated from the neighbouring dates.
 
 ## Input Formats
 
@@ -27,7 +21,7 @@ See `SKILL.md` for what to read and when; this part is procedure, not routing.
 ```
 User: /short-interest GME
 
-Claude retrieves current short float, DTC, borrow rate, and computes Squeeze Score
+The agent requests current short float, DTC and borrow rate through the host data contract, then computes the Squeeze Score
 ```
 
 ### Format 2: Squeeze Potential Scan

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/stock-eval/SKILL.md

@@ -5,6 +5,13 @@ description: Evaluate US stocks with comprehensive fundamental and valuation ana
 
 # US Stock Evaluation
 
+## Evidence request
+
+通过宿主数据契约请求：财务报表（三表与附注，注明期间与 as-of）、行情序列（复权口径与 as-of）、同业可比数据（口径与期间一致）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/stock-eval/references/11-dupont-decomposition.md

@@ -54,7 +54,7 @@ ROE = Tax Burden × Interest Burden × EBIT Margin × Asset Turnover × Equity M
 
 ### Visualization Data Tables
 
-When a visual or report output is needed, populate these tables for chart generation (compatible with `/us-stock-analysis:report-generator`):
+When a visual output is needed, populate these tables for chart generation (the host chart channel renders them; this plugin exports no report):
 
 **Revenue & Earnings Growth**
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/stock-screener/SKILL.md

@@ -5,6 +5,13 @@ description: Screen and rank multiple US stocks across valuation, quality, momen
 
 # Stock Screener
 
+## Evidence request
+
+通过宿主数据契约请求：行情序列（复权口径与 as-of）、财务报表（三表与附注，注明期间与 as-of）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/stock-valuation/SKILL.md

@@ -5,6 +5,13 @@ description: Multi-method stock valuation using DCF, comparable company analysis
 
 # Stock Valuation
 
+## Evidence request
+
+通过宿主数据契约请求：财务报表（三表与附注，注明期间与 as-of）、行情序列（复权口径与 as-of）、同业可比数据（口径与期间一致）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/tax-lens/SKILL.md

@@ -5,6 +5,13 @@ description: US tax mechanics for a position or portfolio — short- vs. long-te
 
 # Tax Lens — What You Actually Keep
 
+## Evidence request
+
+通过宿主数据契约请求：持仓与成本（使用者提供）、使用者提供的输入（成交、持仓或论点记录）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 > ⚠️ **This is not tax advice.** Tax law changes, thresholds are indexed, and the right answer depends on facts this skill cannot see (filing status, state, residency, other income, treaty position). Every output of this skill must open **and** close with: *"Educational illustration of how the rules generally work — confirm with a qualified tax professional before acting."* If the user asks for a filing position, a specific dollar liability to rely on, or anything about a jurisdiction outside the US mechanics described here, say so and stop.
 
 ## ⚠️ Data Verification — Do This Before Any Analysis
@@ -24,12 +31,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-Immediately after the not-tax-advice banner (which is always line one), the output carries this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+Immediately after the not-tax-advice banner (which is always line one), the output carries this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/technical-analysis/SKILL.md

@@ -5,6 +5,13 @@ description: Technical analysis of US stocks using charts and indicators
 
 # Technical Analysis
 
+## Evidence request
+
+通过宿主数据契约请求：行情序列（复权口径与 as-of）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/technical-analysis/references/11-chart-visualization-support.md

@@ -156,7 +156,7 @@ Signals:
 
 ### Integration with Report Generator
 
-When `--chart` flag is used, output is optimized for `/report-generator`:
+When `--chart` flag is used, output is optimized for the host chart channel:
 - All chart data tables formatted for Chart.js/Plotly
 - Color schemes specified for consistent visualization
 - Annotations and overlays clearly defined

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/thesis-tracker/SKILL.md

@@ -5,6 +5,13 @@ description: Write, save, and re-check an investment thesis — KPIs with thresh
 
 # Thesis Tracker — Write It Down, Then Check It
 
+## Evidence request
+
+通过宿主数据契约请求：披露文件（公告、年报或招股书，带回出处与文档日期）、行情序列（复权口径与 as-of）。
+
+缺哪个字段就点名并停止该字段下的分析；用户粘贴的材料标 `pasted by user`。
+
+
 ## ⚠️ Data Verification — Do This Before Any Analysis
 
 Before running any analysis, always retrieve the latest market data for the ticker:
@@ -22,12 +29,12 @@ Never substitute a remembered value for a retrieved one. Getting the number is t
 
 ## 📋 Data & Sources Header — Open Every Output With It
 
-The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard documented on the [Data & Accuracy](https://yennanliu.github.io/InvestSkill/data-and-accuracy.html) page and the first thing `result-validator` looks for:
+The first thing in the output is this provenance block, filled in — never left as placeholders. It is the standard `result-validator` looks for first:
 
 ```
 Data & Sources
   As of:      <date the figures represent, e.g. 2026-06-30>
-  Source:     <primary docs — SEC EDGAR 10-K/10-Q, company IR, FRED, exchange data …>
+  Source:     <venue, document, date — as returned by retrieval>
   Retrieval:  <delegated retrieval | web/tool retrieval | pasted by user>
   Confidence: <HIGH | MEDIUM | LOW>
 ```

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/thesis-tracker/references/10-overview.md

@@ -22,7 +22,7 @@ It fills the gap between Playbook steps 1 (write the thesis), 9 (monitor), and 1
 | **Open** | `thesis-tracker NVDA` (+ pasted analyses, or none) | Interview the user, draft the thesis file, save it to `output/thesis/NVDA.md` |
 | **Update** | `thesis-tracker NVDA --update` (+ current data or a fresh analysis) | Re-read the saved file, refresh every KPI, re-test every trigger, return INTACT / WEAKENED / BROKEN (or INSUFFICIENT EVIDENCE), append to the decision log |
 | **Review** | `thesis-tracker --review` (no ticker) | Read every file in `output/thesis/`, list status and next-check date per ticker, flag anything overdue or sitting at INSUFFICIENT EVIDENCE |
-| **Close** | `thesis-tracker NVDA --close "sold — trigger 2 fired"` | Mark the thesis closed with the exit reason and final P&L; keep the file for `trade-postmortem` |
+| **Close** | `thesis-tracker NVDA --close "sold — trigger 2 fired"` | Mark the thesis closed with the exit reason and final P&L; keep the file for `post-trade-review` |
 
 If the environment cannot write files, output the complete thesis file as a fenced block and tell the user to save it at the stated path — the *contract* matters, not the storage.
 

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/thesis-tracker/references/11-phase-5-the-pre.md

@@ -30,7 +30,7 @@ On every update:
 An INSUFFICIENT EVIDENCE status does not change the file's last confirmed status line; it is recorded in the decision log (`checked — insufficient evidence — <KPIs missing>`) and the next-check date is moved to when the data will exist (usually the next filing).
 
 5. Name **the specific line that changed** — the KPI, its old value, its new value, and the threshold. A status without a line is an opinion.
-6. Append a **decision log** row. The log records what the user *did* and *why*, priced and dated, and names the skill that informed it. It is the raw material for `trade-postmortem`.
+6. Append a **decision log** row. The log records what the user *did* and *why*, priced and dated, and names the skill that informed it. It is the raw material for `post-trade-review`.
 
 **Status is not a signal.** WEAKENED does not mean sell; it means the next check moves up and the position should not be added to. BROKEN means the reason for owning it is gone — what to do about that is a `position-ladder` / tax question, but *adding* is off the table. INSUFFICIENT EVIDENCE means the honest answer is "I don't know yet" — treat it like WEAKENED for position management (no adding) until the data arrives.
 

package/selected-skills/InvestSkill/plugins/us-stock-analysis/skills/thesis-tracker/references/12-example.md

@@ -35,8 +35,8 @@ and appends the log row "2026-08-29 · hold, no add · $131 · margin breach und
 - **Never write the thesis for the user.** Sharpen, question, and structure — but the conviction has to be theirs or the file will not be believed when it says BROKEN.
 - **Hold the KPI count at five.** More KPIs means more noise and a guaranteed breach every quarter; the thesis becomes permanently WEAKENED and the status stops carrying information.
 - **A WEAKENED status blocks adding, not holding.** Pair with `position-ladder`, whose thesis-break gate is meant to read this status.
-- **Closed files are inputs to `trade-postmortem`.** Do not delete them.
-- File paths and the `output/` convention follow `full-report`. Where the environment cannot write, print the file.
+- **Closed files are inputs to `post-trade-review`.** Do not delete them.
+- Keep the `output/` convention described above. Where the environment cannot write, print the file.
 
 ## Thesis Invalidation
 

package/selected-skills/README.md

@@ -6,23 +6,29 @@ automatically, and after the user explicitly enables it in the plugin settings c
 it is visible only to the Capital role(s) declared in `routing.js` (main Agent,
 data analysis, or retrieval/evidence).
 
-The runtime catalog, display tags and provenance live in [`catalog.js`](./catalog.js); role ownership and trigger text live in [`routing.js`](./routing.js). The human-readable 68-row review table is [`ROUTING.md`](./ROUTING.md). Actual loads are recorded programmatically by the role-scoped provider (`skills_used` in the role prompt, `npm run report:skill-usage` for the session-log audit). The
+The runtime catalog, display tags and provenance live in [`catalog.js`](./catalog.js); role ownership and trigger text live in [`routing.js`](./routing.js). The human-readable 70-row review table is [`ROUTING.md`](./ROUTING.md). Actual loads are recorded programmatically by the role-scoped provider (`skills_used` in the role prompt, `npm run report:skill-usage` for the session-log audit). The
 catalog is intentionally broader than any default composition. A candidate may be
 selectable while still carrying a `watchlist` status; that status is a visible signal
 for the user to review the source and decide whether it fits the task.
 
 ## Layout
 
-- `<project>/<skills root>/`: upstream skill files, preserved at a pinned commit. The root
-  is not always `skills/` — it is declared per repository in `SELECTED_SKILL_SKILL_DIRS`
+- `<project>/<skills root>/`: the selectable skill bodies, preserved at a pinned commit. The
+  root is not always `skills/` — it is declared per repository in `SELECTED_SKILL_SKILL_DIRS`
   (`catalog.js`), because upstreams lay skills out differently: one buries them under a
   plugin directory, one groups them two levels deep, and the A-share snapshot has six
   parallel roots, of which exactly one is registered (the line upstream itself calls the
-  source of truth). The newest snapshot needed none of these — its 48 skills sit flat under
-  `skills/`, each frontmatter `name` already matching its directory, and no prose links
-  outside its own folder. Everything else in the snapshot stays unregistered material.
+  source of truth). Only the bodies in `catalog.js` are kept: everything else the upstream
+  shipped — other skills, scripts, MCP servers, evaluation fixtures, docs, sites, CI — was
+  deleted from this repository, not merely excluded from the package. The one exception is
+  `china-stock-research-skills/skills/shared/`, whose reference files the six A-share skills
+  link to directly.
+- `reference-algorithms/`: pure formula implementations moved out of the runtime tree
+  (43 finance `calculate.py` files plus the trading `_lib/calculations.py`). Git-only, never
+  packaged, never executed and never pointed at by a skill body; `README.md` there is the
+  index. They exist as upgrade material for a future computation surface, nothing else.
 - `<project>/UPSTREAM.md`: repository URL, owner, commit, license, runtime scope,
-  maintenance notes, and local changes.
+  maintenance notes, local changes, and the deletion record for everything not kept.
 - `<project>/LICENSE`: original upstream license.
 - `catalog.js`: Capital-owned stable keys, labels, capability tags, and maturity notes.
 
@@ -42,18 +48,21 @@ cadence, scores, benchmark results, and community response guide the status and
 interval; they do not by themselves hide a candidate from the user during this phase.
 The Owner actively looks for useful projects and welcomes pull requests.
 
-This does not grant upstream code new permissions. Upstream scripts are retained as
-source material and are never executed by intake or activation. The Capital provider
-scans only catalog-approved skill directories, registers only explicitly enabled names,
-and does not register the upstream orchestrator that conflicts with Capital routing.
+This does not grant upstream code new permissions. The runtime tree carries Markdown only:
+no scripts, no MCP servers, no evaluation fixtures, no non-catalog skills. Pure formula
+implementations live in `reference-algorithms/`, are never packaged, and are never executed.
+The Capital provider scans only catalog-approved skill directories and registers only
+explicitly enabled names.
 
 Upstream prose **is** edited when it fails the intake checklist — a skill that tells the
-analysis role to run its own web search, to proceed on "training-data estimates", or to
-invoke a helper script would otherwise route around this plugin's data boundaries with no
-tool to stop it. Every such rewrite is mechanical, scoped to catalog-exposed skills, and
-listed per category in that snapshot's `UPSTREAM.md`; `test/selected-skills-content.test.mjs`
-re-checks the result on every run. Do not touch prose for taste, do not rename directories,
-and leave skills that are not in the catalog exactly as upstream wrote them.
+analysis role to run its own web search, to pick a data vendor, to invoke a helper script, or
+to proceed on "training-data estimates" would otherwise route around this plugin's data
+boundaries with no tool to stop it. Where data comes from is stated once, in the
+`capital:data-contract` role-prompt section; each skill only declares what evidence it needs
+(`## Evidence request`) and leaves the retrieval route to the host. Every rewrite is
+mechanical, listed per category in that snapshot's `UPSTREAM.md`, and re-checked by
+`test/selected-skills-content.test.mjs` and `test/selected-skills-structure.test.mjs` on every run.
+Do not touch prose for taste and do not rename directories.
 
 The full checklist lives in `docs/dev/selected-skills.md` §8.6. If a local adaptation is
 required, record it in `UPSTREAM.md`. These are vendored snapshots, not GitHub forks or

package/selected-skills/ROUTING.md

@@ -1,4 +1,4 @@
-# 精选 Skill 路由规划表（68 颗，review 用）
+# 精选 Skill 路由规划表（70 颗，review 用）
 
 > 本表由 `selected-skills/routing.js` 生成，不是第二份真相；角色/触发条件改动以 `routing.js` 为准。
 > 生成规则：`market` 取 catalog 的 A 股/美股标签（都没有则为 global）；`stage`、`methods`、`requires`、
@@ -10,11 +10,11 @@
 
 | 角色 | primaryRole | roles 成员（含多角色） |
 |---|---:|---:|
-| 主 Agent | 29 | 37 |
+| 主 Agent | 31 | 39 |
 | 数据分析 | 36 | 40 |
 | 检索与取证 | 3 | 8 |
 
-## 68 颗
+## 70 颗
 
 | # | name | 中文名 | 角色（primary 在前） | primaryRole | stage | methods | market | 触发条件（zh） | canonicalFor | overlaps 数 |
 |---:|---|---|---|---|---|---|---|---|---|---:|
@@ -86,3 +86,5 @@
 | 66 | `banking-compliance` | 银行资本与流动性指标 | 数据分析 | 数据分析 | research | risk / financial-health | global | 当用户要看银行资本与流动性指标：CET1、RWA、杠杆率、LCR / NSFR 与压力形状时使用；关键词：银行资本、CET1、RWA、LCR、NSFR。 | — | 0 |
 | 67 | `forensic-accounting` | 账务异常与数字分布排查 | 数据分析 | 数据分析 | review | accounting / evidence | global | 当用户要用 Benford、整数重复金额、资金循环与供应商关联方重合作取证式排查时使用；关键词：取证会计、Benford、资金循环、关联方。 | — | 6 |
 | 68 | `cross-market-listing-choice` | 跨市场选线（A/H · ADR/HK） | 主 Agent / 数据分析 | 主 Agent | review | portfolio / tax / risk | a-share / us-stock | 当同一家公司有 A/H、ADR/港股多条上市线，用户要选一条持有、加仓或观察时使用；关键词：A/H 溢价、ADR、选线、汇率、股息税。 | — | 0 |
+| 69 | `macro-event-analysis` | 宏观事件梳理 | 主 Agent | 主 Agent | pre-event | macro / risk | us-stock | 当用户要在未来一到两周的宏观事件里挑出真正影响决策的几项，并看清它们通过利率、汇率、商品与风险偏好传导到持仓时使用；关键词：宏观事件、事件风险、传导、时间窗。 | global/pre-event/macro | 2 |
+| 70 | `market-regime-analysis` | 市场环境判定 | 主 Agent | 主 Agent | monitoring | technical / risk | us-stock | 当用户要按趋势、波动率、广度与事件背景判定当前市场环境，并据此选择趋势跟随、均值回归还是防御时使用；关键词：市场环境、趋势、波动率、广度。 | global/monitoring/regime | 1 |

package/selected-skills/catalog.js

@@ -924,6 +924,44 @@ export const SELECTED_SKILL_CATALOG = Object.freeze([
       en: 'Cold-start experiment; upstream has been quiet since 2026-03-16, so review the sources before use. The hold / trim / tighten / exit call is a method exercise: this plugin holds no broker interface and executes nothing for you. When the original plan or position detail is missing it reviews provisionally and names the gap instead of inventing a tidy prior plan. Options, extended hours and execution constraints follow US / offshore conventions — A-share T+1, daily price limits and the call auction have no counterpart in the text.',
     },
   },
+  {
+    key: 'macroEventAnalysis',
+    name: 'macro-event-analysis',
+    label: { zh: '宏观事件梳理', en: 'Macro Event Analysis' },
+    description: {
+      zh: '把未来一到两周的宏观事件排成决策相关的清单：按市场敏感度、意外空间、传导速度与时间压力排序，再落到利率、汇率、商品与风险偏好的传导路径。',
+      en: 'Turns the next one to two weeks of macro events into a decision-relevant slate, ranked by market sensitivity, surprise capacity, transmission speed and timing pressure, then mapped to rates, FX, commodities and risk appetite.',
+    },
+    owner: 'marian2js',
+    repository: 'trading-skills',
+    status: 'watchlist',
+    lastCommit: '2026-03-16',
+    reviewBy: '2026-11-06',
+    tags: ['trade-discipline', 'macro', 'monitoring', 'us-stock'],
+    riskNote: {
+      zh: '冷启动实验；上游自 2026-03-16 起无提交，启用前请自行复核来源。它排的是事件风险与传导路径，不预测单次数据的方向，也不构成买卖建议；日历缺失时按缺口如实列出，不补一致预期。',
+      en: 'Cold-start experiment; upstream has been quiet since 2026-03-16, so review the sources before use. It ranks event risk and transmission channels, does not predict the direction of a release, and is not investment advice. Missing calendar rows stay missing rather than being filled from memory.',
+    },
+  },
+  {
+    key: 'marketRegimeAnalysis',
+    name: 'market-regime-analysis',
+    label: { zh: '市场环境判定', en: 'Market Regime Analysis' },
+    description: {
+      zh: '按趋势、波动率、广度与事件背景四个锚点保守地判定当前市场环境（健康趋势 / 脆弱趋势 / 过渡 / 防御），并说明哪些证据会推翻这个判定。',
+      en: 'Classifies the current market environment conservatively across trend, volatility, breadth and event backdrop (healthy trend, fragile trend, transition, defensive), and states what evidence would overturn the call.',
+    },
+    owner: 'marian2js',
+    repository: 'trading-skills',
+    status: 'watchlist',
+    lastCommit: '2026-03-16',
+    reviewBy: '2026-11-06',
+    tags: ['trade-discipline', 'technicals', 'risk', 'us-stock'],
+    riskNote: {
+      zh: '冷启动实验；上游自 2026-03-16 起无提交，启用前请自行复核来源。它给的是战术默认值与失效条件，不预测收益、不给置信度百分比，也不代替具体标的的交易计划。',
+      en: 'Cold-start experiment; upstream has been quiet since 2026-03-16, so review the sources before use. It gives tactical defaults and invalidation conditions — no return forecast, no confidence percentages, and no substitute for instrument-level trade planning.',
+    },
+  },
   {
     key: 'chinaEarningsAnalysis',
     name: 'china-earnings-analysis',