Capital Rotation Scanner (US-only)

Scans capital flow across standard US cohorts in one session, identifies rotation direction, classifies winners / losers, names the dominant narrative, and logs a journal file.

Scope: US-only. Do NOT query HK / CN / SG markets (user preference). Sources: Longbridge capital, market-temp. Cite as 长桥证券. Units: 万 USD. Show raw numbers from API; convert to 亿 for headline readability.

When to use

• "今天的资金流向", "板块强弱", "rotation map"
• "卖芯买云", "AI 资金挪去哪了"
• "软件 / 云强不强", "半导体板块今天主力撤了吗"
• End-of-session debrief or pre-Asia-open prep
• Not for single-symbol deep-dive (use longbridge-capital-flow directly)
• Not for live intraday tracking (use market-session-tracker)

Standard cohorts

User watchlist override: read stocks/ directory for symbols the user already tracks; promote those to first-tier in their respective cohort.

Workflow

1. Time check — date + confirm US session state (pre / intraday / post / closed). Adjust analysis date in filename: use the US session date, not Asia local date.

2. Market temperature — single call:

   longbridge market-temp US --format json
   

   Copy

   Report Temperature / Valuation / Sentiment.

3. Index baseline — snapshot mode (gives large/medium/small breakdown):

   longbridge capital SPY.US --format json
   longbridge capital QQQ.US --format json
   

   Copy

   Net large = capital_in.large - capital_out.large. Flag distribution if large net ≪ 0 while small net > 0 (主力—散户背离).

4. Cohort scan — for each cohort, longbridge capital <SYM> --flow --format json | tail -8 to grab the latest cumulative inflow value (the last array element is the running total in 万 USD). Parallelize across symbols.

5. Rotation classification — for each cohort, sum net flows; rank symbols within cohort; identify:

   • Cohort net (sector-level direction)
   • Cohort leader (largest +)
   • Cohort outlier (largest −, especially if cohort net is positive)
   • Cross-cohort rotation: which cohort gained vs lost net flow

6. Narrative identification — pick ONE of:

   • 卖芯买云 (semis − / software +)
   • 卖云买芯 (inverse)
   • 全板派发 (all cohorts negative + indices large-out)
   • 全板吸金 (all positive + indices large-in)
   • 窄口集中 (one cohort dominated by 1-2 names; rest flat)
   • 风险偏好切换 (cyclicals out / defensives in)
   • AI 已变现 vs 未变现 (rank by AI revenue maturity, see §Narrative criteria)

7. Write journal file — ~/git/trade/journal/YYYY-MM-DD-flow.md using the US session date. Use templates/rotation-snapshot.md as scaffold. If the file exists (e.g. re-run same day), append a new section with timestamp; do not overwrite.

8. Surface 3 insights + tomorrow watch — concise, actionable. No vague "市场分化", always name the split.

Distribution detection rules

Use these triggers to label index behavior:

Always state the pattern explicitly; do not say "weak / strong" vaguely.

Narrative criteria — "AI 已变现 vs 未变现"

A common useful narrative axis. Classify cohort flow winners / losers by AI revenue maturity:

• 已变现 (likely to attract flow): MU (HBM revenue confirmed), NOW (AI workflow ARR), ORCL (AI cloud bookings), AMD (MI-series sales), AMZN (AWS Bedrock), PLTR (gov + commercial AIP)
• 未变现 / 纯叙事: SNOW, MRVL (AI guidance hasn't translated), CRWD (AI security narrative), AAPL (Apple Intelligence vague), GOOGL (Gemini monetization weak)

When flow winners cluster in "已变现" and losers in "未变现", call out "narrative 收敛至 AI 已变现窄口" — this is a key macro signal of late-cycle AI selectivity.

CLI quick reference

longbridge market-temp US --format json
longbridge capital SPY.US --format json                    # snapshot (large/med/small)
longbridge capital QQQ.US --flow --format json | tail -8   # time-series cumulative
longbridge capital <SYM>.US --flow --format json | tail -8 # per-symbol

The --flow last-row inflow field is the cumulative net for the session in 万 USD. No date parameter — today's data only.

Failure modes

• Error: request timeout / connect timeout → retry 1-2 times; do not block the report. Mark unavailable symbols with n/a and proceed.
• Cohort scan during US pre-market (04:00–09:30 ET) → data exists but thin; flag report as "pre-market preliminary, not full-session".
• Symbol unavailable on Longbridge (e.g. .SOX.US) → substitute ETF proxy (SMH/SOXX).

Output format (chat reply)

1. Header: market state + session date
2. Index baseline table
3. Cohort tables (one per cohort)
4. Narrative label in bold
5. 3 insights — each with a number and 1-2 sentence claim
6. Tomorrow watch — 4-6 bullets, each with explicit symbol + condition

Tone: match user language (default 文言 for this user, see ~/.claude/CLAUDE.md).

Anti-patterns

• ❌ Querying HK / CN / SG markets (user said US-only)
• ❌ Reporting "市场分化" without naming the split
• ❌ Treating one-symbol-dominated cohort net as broad strength (e.g. semis "+3.4 亿" but MU alone is +4.2 亿)
• ❌ Skipping the journal write step
• ❌ Skipping the narrative label (must pick one)
• ❌ Single-point prediction; use scenario language for tomorrow watch

Related skills

• market-session-tracker — live intraday monitoring of one watchlist
• longbridge-capital-flow — single-symbol drill-down
• longbridge-market-temp — sentiment-only snapshot
• stock-deep-dive — multi-lens single-name research

File layout

capital-rotation/
├── SKILL.md
└── templates/
    └── rotation-snapshot.md