Mm Init

The user may provide a ticker, a company name, or an ambiguous input.

Step 2: Web Search to Verify Identity

Use the WebSearch tool to look up the company:

• If user typed a ticker: search "{INPUT} stock company name exchange sector"
• If user typed a name: search "{INPUT} stock ticker symbol exchange"

From search results, extract: full company name, ticker, exchange, sector/industry.

Step 3: Confirm Company with User

Output the result as text and wait for the user's chat reply:

"I found: {COMPANY_NAME} ({TICKER}) — {EXCHANGE}, {SECTOR}. Is this correct? (yes / or type the correct company name or ticker)"

• If user replies "yes", "y", "correct", "是", "对" etc. → proceed to Step 4
• If user types something else → treat it as a new search query, go back to Step 2
• Repeat until confirmed

Step 4: Ask Report Type

Output as text and wait for reply:

"Report type? (daily / weekly, default: daily)"

• If user replies "weekly" or "w" → set run_mode to weekly
• Any other reply (including empty/enter) → default to daily

Step 5: Enrich via yfinance (Optional)

Try to get additional details:

python3 -c "
import yfinance as yf, json
t = yf.Ticker('TICKER')
info = t.info
print(json.dumps({
    'name': info.get('longName', ''),
    'exchange': info.get('exchange', ''),
    'sector': info.get('sector', ''),
    'industry': info.get('industry', ''),
    'market_cap': info.get('marketCap', 0),
    'currency': info.get('currency', 'USD'),
    'country': info.get('country', ''),
}))
"

If yfinance fails, use the web search results — they are sufficient.

Step 6: Determine Market Profile

Infer from exchange:

• NASDAQ, NYSE, AMEX → US
• HKEX, HKG → HK
• SSE, SZSE → CN
• TSE, JPX → JP
• LSE → UK
• XETRA, Euronext → EU

If unclear, ask the user.

Step 7: Check Existing Workspace

If workspaces/{TICKER}/ already exists:

1. Determine today's trading date (smart trading day logic: pre-market or weekend → use last trading day)
2. Check if a final report exists: look for workspaces/{TICKER}/final/{date}/daily_report.md or weekly_report.md (depends on run_mode in the existing status.json)

Case A: Today's report already exists — ask: "Today's report for {TICKER} ({date}) already exists. Regenerate? (yes / no, default: no)"

• If no → display existing report location and exit
• If yes → reset status.json and continue

Case B: No report for today — auto-reset: set stages_completed: [], stage: "initialized", run_date: "{date}". Continue with workspace creation.

Step 8: Create Directory Tree

mkdir -p workspaces/{TICKER}/{profile,raw,normalized,quant,discussion/{analyst_memos,debate/{round_1,round_2}},drafts,reviews,decision,final,exports/{pdf,web},eval} workspaces/shared/market_context/{raw,normalized,indicators}

Step 9: Generate config.yaml

Read the base config from project root. Prefer config.yaml if it exists (user's local config with real API keys); otherwise fall back to config.example.yaml. Create workspaces/{TICKER}/config.yaml with resolved values:

• company.ticker → confirmed ticker (uppercase)
• company.name → confirmed company name
• company.exchange → resolved exchange
• company.market_profile → inferred market
• company.sector_profile → resolved sector (lowercase, underscores)
• run_mode → user's choice (daily/weekly)

Keep all other values as defaults.

Step 10: Initialize status.json

Write workspaces/{TICKER}/status.json:

{
  "stage": "initialized",
  "started_at": "<ISO timestamp>",
  "updated_at": "<ISO timestamp>",
  "ticker": "<TICKER>",
  "run_mode": "<daily|weekly>",
  "stages_completed": [],
  "current_review_loop": 0,
  "errors": []
}

Step 11: Display Summary

Workspace created for {COMPANY_NAME} ({TICKER})

  Workspace:  workspaces/{TICKER}/
  Config:     workspaces/{TICKER}/config.yaml
  Mode:       {daily|weekly} report
  Market:     {market_profile}
  Sector:     {sector}
  Exchange:   {exchange}

Starting pipeline...

CRITICAL: You MUST use the Skill tool to invoke /mm:run. This is the ONLY allowed way to start the pipeline.

Call the Skill tool with exactly these parameters:

• skill: "mm:run"
• args: "workspaces/{TICKER}"

Rules:

• Do NOT call Skill("mm:mm-orchestrator") or any other skill name — only "mm:run"
• Do NOT execute the orchestrator protocol yourself or bypass /mm:run
• Do NOT skip this step or attempt to "run the pipeline directly"

/mm:run handles progress monitor setup, ownership coordination, and pipeline execution. Bypassing it causes the progress checklist to not appear.

The user does not need to type /mm:run separately. The pipeline runs 14 stages: stages 1-12 execute autonomously, stage 13 (user_review) pauses to collect user feedback on the report, then stage 14 (reflect) runs eval and memory.

Note: This skill's behavior must match plugin/commands/init.md exactly. If the init command flow changes, update both files.

Quality Rules

• Always verify company via web search — never guess
• Present clear "Company Name (TICKER) — Exchange" for confirmation
• Existing workspace: check for today's report before regenerating (see Step 7)
• File content language follows config language field (en or ch). JSON keys always English.