trend-detection

name	trend-detection
description	Detects trend direction, strength, and reversal risk for a DSE stock from OHLCV using moving averages, ADX, price structure (higher highs/lower lows), and multi-timeframe alignment. Use when the user asks "what's the trend", "is GP trending up or down", trend strength, MA crossover, higher-highs/lower-lows read, or reversal warning for a Dhaka Stock Exchange ticker.
license	Apache-2.0
compatibility	Prompt-first Agent Skill. Pure-prompt — no bundled script; the model computes from supplied data.
metadata	{"author":"stock-buddy","version":"0.2.0","prd_refs":["PRD-001:REQ-027"],"mode":["momentum","investment"]}

Trend Detection

Prompt-first, pure-prompt skill. No script is bundled — reason directly over the data using the method below and emit the Thinking Card. Pairs with technical-analysis.

Role & objective

Determine the prevailing trend's direction, strength and reversal risk, returning a score (−1..+1 = down..up), confidence and a rating.

When to use

"What's the trend on GP?", "is this trending up?", "trend strength / ADX read", "any reversal signs?", "MA crossover". Feed the score into signal-synthesizer as a technical input.

Inputs you need

ohlcv — daily bars, oldest-first, ≥50 (more for weekly/monthly context). Optional higher-timeframe series for multi-timeframe alignment.

Method (follow in order)

1. Direction — price vs 50/150/200 SMA; MA stack 50>150>200 (up) or 50<150<200 (down).
2. Structure — higher-highs & higher-lows (uptrend) vs lower-highs & lower-lows (downtrend).
3. Strength — ADX(14): ≥25 strong, 20–25 borderline, <20 weak/range; +DI vs −DI for direction.
4. Multi-timeframe — note daily/weekly/monthly agreement (confluence) or conflict.
5. Reversal risk — price-vs-momentum divergence, volume climax, break of trendline/structure.

Scoring rubric

Combine: direction (±0.4 by MA stack & price), structure (±0.3), strength (×ADX gate: full weight if ADX≥25, half if 20–25, near-zero if <20). Clamp to −1..+1. Rating: ≥0.5 strong_uptrend · ≥0.15 uptrend · −0.15..0.15 sideways · ≤−0.15 downtrend · ≤−0.5 strong_downtrend. Confidence rises with timeframe agreement and ADX; cut it on divergence or <50 bars.

Output (emit this Thinking Card)

{ "skill": "trend-detection", "ticker": "..", "mode": "..", "as_of": "..",
  "score": 0.0, "confidence": 0.0, "rating": "uptrend|sideways|downtrend|..",
  "key_metrics": { "adx_14": 0, "ma_stack": "50>150>200", "structure": "HH/HL",
    "timeframes": {"daily": "..", "weekly": ".."} },
  "reasoning": ["..."], "flags": ["limited_history_<50_bars?", "reversal_divergence?"],
  "disclaimer": "Educational analysis only. Not financial advice." }

DSE pitfalls

• Low ADX (<20) means no real trend — don't read direction into chop; say "sideways".
• Monthly trends on DSE can lag thin trading — weight daily/weekly, note the monthly caveat.
• A reversal signal without volume/structure confirmation is weak — flag, don't over-call.

Optional precision helper

No bundled script — this is a pure-prompt skill. For exact MTF values, use the multi_timeframe skill (deterministic daily/weekly/monthly confluence in momentum mode), or technical-analysis for ADX/MA values.

Worked example

Price above a rising 50>150>200 stack, ADX 30 (+DI leading), HH/HL structure, daily+weekly up, monthly sideways → score ≈ +0.7 → uptrend (strong), confidence ≈ 0.8.

References

See technical-analysis/references/INDICATORS.md for indicator formulas. Output is educational analysis only, never financial advice.