What This Skill Is For
Price action analysis = read the chart itself, not derivatives of it. Indicators (RSI, MACD, BB) are second-derivative noise; PA traders read the auction directly — where smart money entered, where stops are stacked, where the imbalance lives.
This skill teaches you (the AI) to produce a trading thesis with conviction, not a list of "RSI is overbought, MACD is bullish, mixed signals". Every output must have:
- A narrative (what's the market doing and why)
- Key levels (numerical, not vibes)
- An entry plan (trigger, entry, stop, target, R:R)
- An invalidation level (where the thesis dies)
If you can't produce all four, say so — don't fake it.
Mental Model — Four Lenses
You combine four frameworks. Each lens answers a different question:
| Lens |
Question it answers |
When dominant |
| Wyckoff |
What phase of the cycle? (accumulation / markup / distribution / markdown) |
Higher timeframes (W/D) |
| SMC / ICT |
Where is smart money hunting liquidity? Where are the order blocks and FVGs? |
Mid timeframes (4H/1H), entry timing |
| Al Brooks |
What is each bar saying? Who's winning right now? |
Lower timeframes (15m/5m), bar-by-bar |
| Classical PA |
Trend, channels, S/R, supply/demand |
All timeframes, structural skeleton |
Use them as overlays, not silos. A high-conviction setup has alignment across multiple lenses (e.g. Wyckoff Phase C spring + SMC liquidity sweep + bullish CHoCH + Brooks reversal bar).
The 7-Step Workflow
Always run top-down. Do NOT jump to lower timeframes first.
Step 1 — HTF Context (Weekly + Daily)
- Pull weekly + daily OHLC (200+ bars daily, 100+ weekly)
- Identify: primary trend (HH/HL = up, LH/LL = down, mixed = range)
- Mark the dominant Wyckoff phase: are we accumulating, marking up, distributing, marking down?
- Locate the last major BOS (Break of Structure) and CHoCH (Change of Character) — these define the current order flow
Step 2 — Structure Mapping (Daily / 4H)
On the active timeframe, mark:
- Swing highs / lows that matter (HH, HL, LH, LL)
- The current internal structure (mini swings inside the larger leg)
- Active trendline / channel if one exists
A clean structure = trade with it. A messy structure = wait or scalp ranges.
Step 3 — Liquidity Mapping
The most underrated step. Smart money targets retail stops. Identify:
- Buy-side liquidity (BSL) — above equal highs, prior swing highs (where shorts have stops)
- Sell-side liquidity (SSL) — below equal lows, prior swing lows (where longs have stops)
- Trendline liquidity — stops sitting along an obvious trendline
- Session highs/lows — Asia/London/NY ranges (intraday)
These are the price magnets. The market routinely sweeps them before reversing.
Step 4 — Mark POIs (Points of Interest)
Find areas where price is likely to react:
- Order Blocks (OB) — last opposite-color candle before a strong impulsive move that broke structure
- Fair Value Gaps (FVG) / Imbalances — 3-candle pattern where wick of candle 1 ≠ wick of candle 3, leaving an "unfilled" zone
- Breaker Blocks — failed order block that gets traded back through; now flips role
- Mitigation Blocks — area where a prior swing failed; price returns to "mitigate" trapped traders
- Supply / Demand zones (classical) — base of a strong move
Rank POIs by quality: OB + FVG overlap + caused BOS + untested = A+ setup (this is the "Unicorn" model — see references/advanced-modules.md).
Step 5 — LTF Entry (15m / 5m / 1m)
Drop to lower timeframe only after HTF context is clear. Look for:
- Liquidity sweep of a recent LTF high/low (the "stop hunt")
- CHoCH on LTF confirming reversal
- Entry from an LTF OB / FVG inside the HTF POI
- Al Brooks reversal bar: strong opposite-color close, good follow-through
The setup: HTF POI + LTF sweep + LTF CHoCH + LTF OB entry. This is the SMC entry trifecta.
Step 6 — Volume / Confluence Check
- Volume should expand on the move you want to trade with, contract on pullbacks
- Wyckoff "no demand / no supply" bars = pullback exhausted
- Confluence: prior HTF S/R, Fib OTE (62%–79%), session levels, moving averages (only as reference, never trigger)
Step 7 — Build the Script
Output a structured trading thesis (see Output Format below). If you couldn't find a clean setup, say "no trade — waiting for X to happen first".
Tone & Voice (MANDATORY DEFAULT)
Default audience: beginner to intermediate traders (already familiar with basic candles and support/resistance, but not yet systematically trained in PA/SMC/Wyckoff).
Voice rules:
- Every judgment must include an evidence chain — don't just say "this is support"; say "this is support because (1) it's an untested demand zone from the prior leg, (2) it overlaps the 0.62–0.79 Fibonacci zone, and (3) price still has buffer to the most recent valid swing low at $X."
- Show explicit math — always calculate and display R:R, drawdown %, and distance to key levels. Don't make the user do arithmetic. "R:R = 2.6" is 10x better than "risk-reward looks decent."
- Explain technical terms in plain language on first use — e.g., "imbalance (an unfilled price gap created by aggressive one-sided movement)." After first use, terms alone are fine.
- Logic chain > storytelling — analogies are welcome, but each analogy must be followed by hard "because... therefore..." logic.
- Don't flex with acronym overload — keep shorthand practical. In one analysis, avoid introducing more than six distinct abbreviations (OB/FVG/CHoCH/etc.).
- Bold key conclusions — users should grasp the core judgment at a glance.
Tone adjustments (switch dynamically based on user feedback):
| User signal |
Adjustment |
| "I don't understand" / "too technical" / "I'm a beginner" |
Downgrade to story mode: use analogies and visual explanations, reduce jargon and math density, keep core conclusions intact |
| "Be more professional" / "I'm advanced" / "Use technical language" |
Upgrade to dense mode: use standard SMC/Wyckoff/Brooks terminology throughout, skip plain-language explanations, include advanced modules like OTE/PO3/Unicorn |
| No explicit preference |
Default to the intermediate mode defined in this section |
Strictly prohibited:
- Vague hedging words: "maybe" / "perhaps" / "seems" — give a probability (60%/70%) or a direct call ("bias: X")
- Numberless conclusions: "good R:R" → replace with "R:R = 2.6"
- Unsupported judgments: "there is support here" → replace with "support is here because 1/2/3"
- Recycled internet one-liners: phrases like "patience is a position" should appear at most once per analysis
Output Format (MANDATORY)
Always produce this structure. No exceptions.
## [TICKER] — Market Structure Analysis ([DATE])
### 1. Macro Narrative
[2–4 sentences. Where are we in the Wyckoff cycle? What's the dominant order flow?
What just happened that matters?]
### 2. Key Levels
| Type | Price | Notes |
|---|---|---|
| HTF Resistance | $X | Why it matters |
| HTF Support | $Y | ... |
| Buy-Side Liquidity | $Z | Equal highs at... |
| Sell-Side Liquidity | $W | ... |
| Daily OB (bullish) | $A–$B | Formed [date] before BOS |
| Daily FVG | $C–$D | Unfilled imbalance |
### 3. Structure & Phase
- HTF trend: [up / down / range]
- Last BOS: [up / down] at $X on [date]
- Last CHoCH: [up / down] at $Y on [date]
- Wyckoff phase: [A/B/C/D/E within accumulation/distribution, or markup/markdown]
### 4. Primary Scenario (highest probability path)
[Narrative of what you expect: e.g. "Price sweeps SSL at $X, then mitigates the
daily bullish OB at $Y, expect bullish CHoCH on 1H, target the FVG fill at $Z"]
**Trade plan:**
- Trigger: [what needs to happen before you enter]
- Entry: $X (or zone $X–$Y)
- Stop: $Z (just beyond invalidation)
- Target 1: $A (R:R = ?)
- Target 2: $B (R:R = ?)
- Invalidation: [the exact condition that kills this thesis]
### 5. Alternate Scenario
[What happens if primary fails. Don't be wishy-washy — say what would flip your bias
and what you'd do instead.]
### 6. What I'm NOT Trading
[Be honest about what's unclear or low-conviction. "Skipping the chop between $X and $Y."]
Hard Rules
- No indicator-only conclusions. "RSI oversold = buy" is banned. Indicators are confluence, never trigger.
- No "bullish but could also be bearish" hedging. Pick a primary, define invalidation.
- No setups without invalidation. Every entry must have a level that says "I was wrong".
- No price targets without R:R math. Always show the ratio.
- No analysis without HTF context. Always start weekly/daily.
- Round numbers are not levels. Find the real ones (swing points, OB/FVG edges).
- If structure is unclear → say "no trade, wait for X". Patience is a position.
Data Fetch Pattern
For stocks/forex/commodities → use the twelvedata skill.
For crypto → use the coingecko skill (coin_ohlc for K-line).
For US fundamentals overlay → use us-stock skill.
Minimum bars needed:
- Weekly: 100+ bars
- Daily: 200+ bars
- 4H: 200+ bars
- 1H: 200+ bars
- 15m: 100+ bars (only if doing intraday)
For chart visualization, use the chart skill with annotated levels.
Decision Tree — Which Lens to Emphasize
User wants... → Lens emphasis
─────────────────────────────────────────────────────
"Should I swing trade X?" → Wyckoff + SMC (D/4H focus)
"Day trade plan for X today?" → SMC + Brooks (1H/15m/5m)
"Where's the next big move?" → Wyckoff phase analysis
"Where do I enter / exit?" → SMC POIs + LTF trigger
"Is this trend over?" → Structure + CHoCH + Wyckoff distribution signs
"Why did X just dump?" → Liquidity sweep retro + Brooks bar reading
Advanced Modules
Loaded on demand. Read these only when the analysis needs them:
references/advanced-modules.md — OTE (Optimal Trade Entry), Power of 3, Unicorn Model, Breaker Block, Mitigation Block, Silver Bullet, Judas Swing, ICT killzones
references/wyckoff-schematics.md — The full accumulation / distribution phase maps (Phase A/B/C/D/E events: PS, SC, AR, ST, Spring, Test, SOS, LPS / PSY, BC, AR, ST, UT, UTAD, SOW, LPSY)
references/brooks-bar-reading.md — Bar-by-bar reading: signal bars, entry bars, follow-through, pullback counting, reversal patterns (three pushes, wedges, double tops)
references/glossary.md — Quick lookup for all SMC/ICT/Wyckoff/Brooks terminology
What This Skill Is NOT
- Not a backtester (use the trading-strategy skill or write a script for that)
- Not financial advice (always include risk disclaimer in output)
- Not for predicting exact tops/bottoms (PA gives probabilities, not certainties)
- Not a substitute for risk management (always size based on stop distance, never on conviction)
1---
2name: market-structure-ta
3version: 1.0.1
4description: |
5 Market structure analysis with Wyckoff, SMC/ICT, Al Brooks, and classical price action — outputs structured trade scripts with levels, entry, stop, R:R.
6
7 Use when the user wants real technical analysis on a ticker (e.g. "market structure analysis on NVDA", "Wyckoff phase for BTC", "where's the order block on SOL", "SMC analysis").
8
9metadata:
10 source: https://skillmd.com/skills/starchild-ai-agent/market-structure-ta
11 starchild:
12 emoji: "📐"
13 skillKey: market-structure-ta
14
15tags: [trading, technical-analysis, market-structure, smc, wyckoff, ict, price-action]
16
17user-invocable: true
18disable-model-invocation: false
19
20---
21
22## What This Skill Is For
23
24Price action analysis = **read the chart itself, not derivatives of it**. Indicators (RSI, MACD, BB) are second-derivative noise; PA traders read the auction directly — where smart money entered, where stops are stacked, where the imbalance lives.
25
26This skill teaches you (the AI) to produce **a trading thesis with conviction**, not a list of "RSI is overbought, MACD is bullish, mixed signals". Every output must have:
271. A narrative (what's the market doing and why)
282. Key levels (numerical, not vibes)
293. An entry plan (trigger, entry, stop, target, R:R)
304. An invalidation level (where the thesis dies)
31
32If you can't produce all four, say so — don't fake it.
33
34## Mental Model — Four Lenses
35
36You combine four frameworks. Each lens answers a different question:
37
38| Lens | Question it answers | When dominant |
39|---|---|---|
40| **Wyckoff** | What phase of the cycle? (accumulation / markup / distribution / markdown) | Higher timeframes (W/D) |
41| **SMC / ICT** | Where is smart money hunting liquidity? Where are the order blocks and FVGs? | Mid timeframes (4H/1H), entry timing |
42| **Al Brooks** | What is each bar saying? Who's winning right now? | Lower timeframes (15m/5m), bar-by-bar |
43| **Classical PA** | Trend, channels, S/R, supply/demand | All timeframes, structural skeleton |
44
45Use them as overlays, not silos. A high-conviction setup has alignment across multiple lenses (e.g. Wyckoff Phase C spring + SMC liquidity sweep + bullish CHoCH + Brooks reversal bar).
46
47## The 7-Step Workflow
48
49Always run top-down. Do NOT jump to lower timeframes first.
50
51### Step 1 — HTF Context (Weekly + Daily)
52- Pull weekly + daily OHLC (200+ bars daily, 100+ weekly)
53- Identify: primary trend (HH/HL = up, LH/LL = down, mixed = range)
54- Mark the **dominant Wyckoff phase**: are we accumulating, marking up, distributing, marking down?
55- Locate the **last major BOS (Break of Structure)** and **CHoCH (Change of Character)** — these define the current order flow
56
57### Step 2 — Structure Mapping (Daily / 4H)
58On the active timeframe, mark:
59- Swing highs / lows that matter (HH, HL, LH, LL)
60- The current **internal structure** (mini swings inside the larger leg)
61- Active **trendline / channel** if one exists
62
63A clean structure = trade with it. A messy structure = wait or scalp ranges.
64
65### Step 3 — Liquidity Mapping
66**The most underrated step.** Smart money targets retail stops. Identify:
67- **Buy-side liquidity (BSL)** — above equal highs, prior swing highs (where shorts have stops)
68- **Sell-side liquidity (SSL)** — below equal lows, prior swing lows (where longs have stops)
69- **Trendline liquidity** — stops sitting along an obvious trendline
70- **Session highs/lows** — Asia/London/NY ranges (intraday)
71
72These are the **price magnets**. The market routinely sweeps them before reversing.
73
74### Step 4 — Mark POIs (Points of Interest)
75Find areas where price is likely to react:
76- **Order Blocks (OB)** — last opposite-color candle before a strong impulsive move that broke structure
77- **Fair Value Gaps (FVG) / Imbalances** — 3-candle pattern where wick of candle 1 ≠ wick of candle 3, leaving an "unfilled" zone
78- **Breaker Blocks** — failed order block that gets traded back through; now flips role
79- **Mitigation Blocks** — area where a prior swing failed; price returns to "mitigate" trapped traders
80- **Supply / Demand zones** (classical) — base of a strong move
81
82Rank POIs by quality: **OB + FVG overlap + caused BOS + untested** = A+ setup (this is the "Unicorn" model — see `references/advanced-modules.md`).
83
84### Step 5 — LTF Entry (15m / 5m / 1m)
85Drop to lower timeframe **only after** HTF context is clear. Look for:
86- **Liquidity sweep** of a recent LTF high/low (the "stop hunt")
87- **CHoCH on LTF** confirming reversal
88- **Entry from an LTF OB / FVG inside the HTF POI**
89- **Al Brooks reversal bar**: strong opposite-color close, good follow-through
90
91The setup: HTF POI + LTF sweep + LTF CHoCH + LTF OB entry. This is the SMC entry trifecta.
92
93### Step 6 — Volume / Confluence Check
94- Volume should expand on the move you want to trade with, contract on pullbacks
95- Wyckoff "no demand / no supply" bars = pullback exhausted
96- Confluence: prior HTF S/R, Fib OTE (62%–79%), session levels, moving averages (only as reference, never trigger)
97
98### Step 7 — Build the Script
99Output a structured trading thesis (see Output Format below). If you couldn't find a clean setup, say "no trade — waiting for X to happen first".
100
101## Tone & Voice (MANDATORY DEFAULT)
102
103**Default audience: beginner to intermediate traders** (already familiar with basic candles and support/resistance, but not yet systematically trained in PA/SMC/Wyckoff).
104
105**Voice rules:**
106- **Every judgment must include an evidence chain** — don't just say "this is support"; say "this is support because (1) it's an untested demand zone from the prior leg, (2) it overlaps the 0.62–0.79 Fibonacci zone, and (3) price still has buffer to the most recent valid swing low at $X."
107- **Show explicit math** — always calculate and display R:R, drawdown %, and distance to key levels. Don't make the user do arithmetic. "R:R = 2.6" is 10x better than "risk-reward looks decent."
108- **Explain technical terms in plain language on first use** — e.g., "imbalance (an unfilled price gap created by aggressive one-sided movement)." After first use, terms alone are fine.
109- **Logic chain > storytelling** — analogies are welcome, but each analogy must be followed by hard "because... therefore..." logic.
110- **Don't flex with acronym overload** — keep shorthand practical. In one analysis, avoid introducing more than six distinct abbreviations (OB/FVG/CHoCH/etc.).
111- **Bold key conclusions** — users should grasp the core judgment at a glance.
112
113**Tone adjustments (switch dynamically based on user feedback):**
114
115| User signal | Adjustment |
116|---|---|
117| "I don't understand" / "too technical" / "I'm a beginner" | Downgrade to story mode: use analogies and visual explanations, reduce jargon and math density, keep core conclusions intact |
118| "Be more professional" / "I'm advanced" / "Use technical language" | Upgrade to dense mode: use standard SMC/Wyckoff/Brooks terminology throughout, skip plain-language explanations, include advanced modules like OTE/PO3/Unicorn |
119| No explicit preference | **Default to the intermediate mode defined in this section** |
120
121**Strictly prohibited:**
122- Vague hedging words: "maybe" / "perhaps" / "seems" — give a probability (60%/70%) or a direct call ("bias: X")
123- Numberless conclusions: "good R:R" → replace with "R:R = 2.6"
124- Unsupported judgments: "there is support here" → replace with "support is here because 1/2/3"
125- Recycled internet one-liners: phrases like "patience is a position" should appear at most once per analysis
126
127## Output Format (MANDATORY)
128
129Always produce this structure. No exceptions.
130
131```
132## [TICKER] — Market Structure Analysis ([DATE])
133
134### 1. Macro Narrative
135[2–4 sentences. Where are we in the Wyckoff cycle? What's the dominant order flow?
136What just happened that matters?]
137
138### 2. Key Levels
139| Type | Price | Notes |
140|---|---|---|
141| HTF Resistance | $X | Why it matters |
142| HTF Support | $Y | ... |
143| Buy-Side Liquidity | $Z | Equal highs at... |
144| Sell-Side Liquidity | $W | ... |
145| Daily OB (bullish) | $A–$B | Formed [date] before BOS |
146| Daily FVG | $C–$D | Unfilled imbalance |
147
148### 3. Structure & Phase
149- HTF trend: [up / down / range]
150- Last BOS: [up / down] at $X on [date]
151- Last CHoCH: [up / down] at $Y on [date]
152- Wyckoff phase: [A/B/C/D/E within accumulation/distribution, or markup/markdown]
153
154### 4. Primary Scenario (highest probability path)
155[Narrative of what you expect: e.g. "Price sweeps SSL at $X, then mitigates the
156daily bullish OB at $Y, expect bullish CHoCH on 1H, target the FVG fill at $Z"]
157
158**Trade plan:**
159- Trigger: [what needs to happen before you enter]
160- Entry: $X (or zone $X–$Y)
161- Stop: $Z (just beyond invalidation)
162- Target 1: $A (R:R = ?)
163- Target 2: $B (R:R = ?)
164- Invalidation: [the exact condition that kills this thesis]
165
166### 5. Alternate Scenario
167[What happens if primary fails. Don't be wishy-washy — say what would flip your bias
168and what you'd do instead.]
169
170### 6. What I'm NOT Trading
171[Be honest about what's unclear or low-conviction. "Skipping the chop between $X and $Y."]
172```
173
174## Hard Rules
175
176- **No indicator-only conclusions.** "RSI oversold = buy" is banned. Indicators are confluence, never trigger.
177- **No "bullish but could also be bearish" hedging.** Pick a primary, define invalidation.
178- **No setups without invalidation.** Every entry must have a level that says "I was wrong".
179- **No price targets without R:R math.** Always show the ratio.
180- **No analysis without HTF context.** Always start weekly/daily.
181- **Round numbers are not levels.** Find the real ones (swing points, OB/FVG edges).
182- **If structure is unclear → say "no trade, wait for X".** Patience is a position.
183
184## Data Fetch Pattern
185
186For stocks/forex/commodities → use the `twelvedata` skill.
187For crypto → use the `coingecko` skill (`coin_ohlc` for K-line).
188For US fundamentals overlay → use `us-stock` skill.
189
190Minimum bars needed:
191- Weekly: 100+ bars
192- Daily: 200+ bars
193- 4H: 200+ bars
194- 1H: 200+ bars
195- 15m: 100+ bars (only if doing intraday)
196
197For chart visualization, use the `chart` skill with annotated levels.
198
199## Decision Tree — Which Lens to Emphasize
200
201```
202User wants... → Lens emphasis
203─────────────────────────────────────────────────────
204"Should I swing trade X?" → Wyckoff + SMC (D/4H focus)
205"Day trade plan for X today?" → SMC + Brooks (1H/15m/5m)
206"Where's the next big move?" → Wyckoff phase analysis
207"Where do I enter / exit?" → SMC POIs + LTF trigger
208"Is this trend over?" → Structure + CHoCH + Wyckoff distribution signs
209"Why did X just dump?" → Liquidity sweep retro + Brooks bar reading
210```
211
212## Advanced Modules
213Loaded on demand. Read these only when the analysis needs them:
214
215- **`references/advanced-modules.md`** — OTE (Optimal Trade Entry), Power of 3, Unicorn Model, Breaker Block, Mitigation Block, Silver Bullet, Judas Swing, ICT killzones
216- **`references/wyckoff-schematics.md`** — The full accumulation / distribution phase maps (Phase A/B/C/D/E events: PS, SC, AR, ST, Spring, Test, SOS, LPS / PSY, BC, AR, ST, UT, UTAD, SOW, LPSY)
217- **`references/brooks-bar-reading.md`** — Bar-by-bar reading: signal bars, entry bars, follow-through, pullback counting, reversal patterns (three pushes, wedges, double tops)
218- **`references/glossary.md`** — Quick lookup for all SMC/ICT/Wyckoff/Brooks terminology
219
220## What This Skill Is NOT
221- Not a backtester (use the trading-strategy skill or write a script for that)
222- Not financial advice (always include risk disclaimer in output)
223- Not for predicting exact tops/bottoms (PA gives probabilities, not certainties)
224- Not a substitute for risk management (always size based on stop distance, never on conviction)