himanshuj16/tradefloor/analyze

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Run the full tradefloor desk on one stock, index, ETF, future, FX pair or crypto asset on any exchange - analysts, bull/bear debate, research manager, trader, risk committee, portfolio manager - and return the expected direction with probabilities, a sized trade plan and the reasoning. Use when the user asks where a stock or index is heading, whether to buy or sell it, or for a trade idea on it.

Install

npx vskill@latest install himanshuj16/tradefloor/analyze

SKILL.md

tradefloor: full desk analysis

Arguments: $ARGUMENTS (if that shows literally, take the request from the user's message).

You are the desk coordinator. You do not analyse; you route work to the roles, keep the run folder in order, and deliver the portfolio manager's decision. Every role writes a file and later roles read those files: the run folder is the shared state.

How roles run in this host

Delegate each role to a subagent if your host has them. The subagent's name is tradefloor:<role> in Claude Code and Devin, tradefloor-<role> in OpenCode, and <role> in Gemini CLI, Copilot CLI and Cursor. Launch the roles marked parallel together when the host allows it.

If the host has no subagents, or the named one is not available, play the role yourself: read references/<role>.md in this skill's folder, follow it exactly, write its file, then continue. Finish one role completely before starting the next, and when playing a role use only what that role is allowed to read. A bull that has already seen the bear's case is a worse bull.

Every role receives the same header of facts: SYMBOL, NAME, EXCHANGE, CURRENCY, KIND, AS_OF, HORIZON, HORIZON_DAYS, RUN_DIR, IS_HISTORICAL, PROXIES, and the price unit note if any.

1. Parse the request

  • INSTRUMENT: anything that is not one of the tokens below. A ticker (RELIANCE,

7203.T, NSE:TCS), a company name, or an index name (nifty 50, s&p 500, dax).

  • HORIZON: 1w, 2w, 1m, 3m, 6m, 1y. Default: defaults.default_horizon.
  • AS_OF: a YYYY-MM-DD date. Absent means today. A past date is a historical run:

IS_HISTORICAL=true.

  • DEPTH: quick, standard (default) or deep. quick runs the direction

procedure instead of the desk.

  • SHORT_OK: true if the user wrote short-ok or said they can short.

2. Resolve

Call resolve_symbol(INSTRUMENT). It returns the resolution and defaults (the user's settings). On candidates, pick the match for the user's words and home market; if two listings are plausible, ask. Then call direction_forecast(symbol, HORIZON, as_of) once to learn HORIZON_DAYS.

Call run_begin(kind="analyze", request=<parsed request object>). Use its workspace as RUN_DIR. Write RUN_DIR/00-request.md with the parsed request, the resolution, DEPTH and ROUNDS (below).

3. Analysts (parallel)

market-analyst, fundamentals-analyst, news-analyst, sentiment-analyst. If one fails, retry once; then write its file as "no report: <error>" and continue.

4. Debate

ROUNDS = defaults.debate_rounds for standard, plus one for deep. Round 1: bull-researcher and bear-researcher in parallel, ROUND=1. Later rounds: both again, each given the other's previous file. Then research-manager with the list of debate files.

5. Trader

trader, told whether SHORT_OK.

6. Risk committee (parallel)

risk-aggressive, risk-neutral, risk-conservative, ROUND=1. Deep runs add a second round in which each seat reads the other two seats' round-1 files.

7. Portfolio manager

portfolio-manager with DEPTH. It writes 50-decision.md and records the call with journal_record.

8. Deliver

Reply, under 40 lines:

  1. SYMBOL (NAME) · EXCHANGE · CURRENCY · as of DATE · horizon HORIZON
  2. Direction, confidence, up/down/flat probabilities beside the quant prior, and the

tilt. Low confidence with no tilt: say the evidence shows no edge.

  1. Expected 80% range.
  2. Rating and plan: action, entry zone, stop, targets, size, time stop; the proxy for

an index.

  1. Why and Against, up to three bullets each.
  2. Invalidation.
  3. The run folder and journal id; the review procedure scores the call after HORIZON.
  4. Not investment advice. Public data, possibly delayed; verify before acting.

Rules

  • No analysis of your own between steps. Roles read files, not your summaries of them.
  • Every number in the final reply comes from 50-decision.md or 01-market.md.
  • If resolving or the price history fails, stop and say what failed.

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
5

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/himanshuj16/tradefloor/analyze/badge)](https://verified-skill.com/skills/himanshuj16/tradefloor/analyze)

Something wrong with this skill or its page? Tell us.