analyze-report

Files (2)

• SKILL.md
• twelvedata_1h.sh · scripts/

SKILL.md (first lines)

# Stock Multi-Timeframe Entry & Change Report

## Purpose

Produce a technical-analysis write-up for a single stock: a real-time snapshot,
a multi-timeframe read (1h, daily, weekly) on whether current
conditions look like a reasonable entry, and a summary of what's changed since
last week, last quarter, and last year. This skill produces **informational
technical analysis**, not a trade recommendation — the framing and disclaimers
below are required, not optional.

## Invocation

Accepts a ticker or a company name, e.g. `Analyze-Report AAPL`,
`Analyze-Report Microsoft`. If a company name is given, resolve it to a
ticker first (step 0) rather than guessing.

## Configuration

The 1h timeframe uses Twelve Data instead of the primary market-data tools
(their intraday endpoints are gated on this connector's tier — see Known
limitations). It requires a free Twelve Data API key, stored as
`TWELVE_DATA_API_KEY` in a `.env` file at the plugin root — **never commit
this file** (it's covered by `.gitignore`). Run
`scripts/twelvedata_1h.sh <SYMBOL>` to fetch it; if the key is missing or the
script fails, skip the 1h timeframe, note it as unavailable in the report,
and continue with daily + weekly.

## Language

Write the final output in Vietnamese, regardless of what language the
request was in. Keep ticker symbols, indicator names (RSI, MACD, EMA/SMA),
and numeric price levels in their standard form — translate the surrounding
prose, headers, and explanations into Vietnamese.

## Data source discipline

Prefer the connected market-data tools (Alpha Vantage-backed, plus Twelve
Data for 1h — see Configuration) over web search — they return exact
computed values instead of scraped/approximate ones, and don't have the
reliability problems of pulling numbers out of JS-rendered dashboards.
Request `datatype: "json"` on every Alpha Vantage call (the tools default to
CSV, and JSON is far less error-prone to parse). Use web search/fetch only
for things the data tools don't cover: same-day macro catalysts, sector
context, or when a tool call fails.

This skill analyzes price/volume data and computed indicator values — it
does **not** relay other people's trading opinions, calls, or predictions.
`NEWS_SENTIMENT` returns aggregated sentiment computed from articles, which
is fine to use as a data point; individual traders'/analysts' subjective
takes, forum posts, or "my outlook" commentary are not.

## Workflow

### 0. Resolve the ticker

If the input isn't already an exact ticker, call `SYMBOL_SEARCH` with the
given name as `keywords`. Prefer the best match with `region: United States` and
`type: Equity` unless the user specified a different market. If multiple
plausible matches exist and it's genuinely ambiguous, ask which one instead
of guessing.

### 1. Real-time snapshot

- `GLOBAL_QUOTE` for current price, day's OHLC, volume, and change vs. prior
  close. Try `entitlement: "realtime"` first; if the response indicates
  delayed data, state that plainly in the report rather than presenting it
  as live.
- `COMPANY_OVERVIEW` for lightweight context: sector, industry, market cap.
  One call, used only to frame the report — don't turn this into a
  fundamentals writeup.
- `EARNINGS_CALENDAR` (symbol-scoped, horizon `3month`) — if an earnings
  date falls within the next ~2 weeks, flag it explicitly in the report as
  a factor that can override the technical picture regardless of chart
  setup.

### 2. Per-timeframe technical read

**Daily and weekly** (primary market-data tools):
- Price series: `TIME_SERIES_DAILY` (outputsize `compact` is enough for
  trend/level context — `_ADJUSTED` is gated on this connector) and
  `TIME_SERIES_WEEKLY_ADJUSTED`.
- `RSI` (`time_period: 14`, `series_type: close`) at `daily` and `weekly`.
- `OBV` and `MFI` (`time_period: 14`) at `daily` and `weekly`, for
  volume/money-flow confirmation.
- `MACD` is unreliable on this connector (fails at every interval on the
  current tier) — treat it as best-effort: try it, but don't block the
  report or call out its absence if it fails.

**1h** (Twelve Data — see Configuration):
- Run `scripts/twelvedata_1h.sh <SYMBOL>`, which returns `time_series`,
  `rsi`, `mfi`, and `obv` at `1h` in one call.
- If the script fails (missing key, rate limit, etc.), skip this timeframe
  entirely rather than guessing values — note it as unavailable in the
  final report.

Note per timeframe:
- **Trend**: up/down/sideways, and price relative to recent price structure
  (use SMA/EMA 20 and 50 if you also pull those — optional, don't add calls
  just for this if the price series already makes the trend clear).
- **Key levels**: nearest support/resistance from recent price structure.
- **Momentum**: RSI reading (>70 overbought / <30 oversold as reference,
  not a rule), MACD line vs. signal line.
- **Money flow**: OBV/MFI confirming the move (rising on advances) or
  diverging from it (price up, flow weakening).

### 3. Compare timeframes to find the entry

The goal is to find where a good entry point exists, not to require every
timeframe to agree:
- Evaluate each timeframe's setup independently first — a clean setup on
  one timeframe (e.g. 1h pulling into a well-defended support with
  rising money flow) is a valid finding even if daily/weekly don't show the
  same clarity.
- Then note where timeframes reinforce or conflict (e.g. "good short-term
  1h entry, but daily trend still down — favors a short hold over a
  swing"). Lack of full alignment doesn't by itself disqualify an entry;
  state plainly which timeframe(s) the case is strongest on and why.
- Is price near a key level? Is momentum/money flow confirming or

From quyenanh198/Skills, MIT.

More from quyenanh198/Skills