Check a dataset before analysis
Find missing values, inconsistent formats, duplicates, and questions that could invalidate results.
Detect and classify candlestick patterns from ingested OHLCV data
---
name: market-pattern-4aee6964ffef
description: Detect and classify candlestick patterns from ingested OHLCV data
argument-hint: "<symbol> [--period 1D]"
allowed-tools: mcp__claude-flow__memory_search mcp__claude-flow__memory_list mcp__claude-flow__memory_store mcp__claude-flow__agentdb_pattern-store mcp__claude-flow__agentdb_pattern-search mcp__claude-flow__ruvllm_hnsw_route Bash
---
# Market Pattern
Scan ingested OHLCV data for known candlestick patterns, classify them by type and reliability, and store for future reference.
## When to use
When you need to identify candlestick patterns (doji, hammer, engulfing, head-shoulders, etc.) in market data. Requires data to be ingested first via `market-ingest`.
## Steps
1. **Load candles** -- call `mcp__claude-flow__memory_search` (or `memory_list`) on the `market-data` namespace to retrieve normalized OHLCV data for the symbol and period. The `memory_*` tool family routes by namespace; the `agentdb_hierarchical-*` family does NOT (it routes by tier), so use `memory_*` for namespaced reads.
2. **Scan for patterns** -- iterate through candle sequences looking for:
- Single-candle: doji (open ~= close), hammer (long lower wick), inverted hammer
- Two-candle: bullish/bearish engulfing
- Three-candle: morning star, evening star, three white soldiers, three black crows
- Multi-candle: head & shoulders, double top/bottom, cup & handle
3. **Classify** -- for each detection, assign: pattern name, type (reversal/continuation), direction (bullish/bearish), reliability score (0.0-1.0)
4. **Rank** -- sort by reliability score descending
5. **Store** -- two paths (per ruflo-cost-tracker ADR-0001 dual-path pattern):
- **Pattern-store (typed, recommended)**: `mcp__claude-flow__agentdb_pattern-store` with `type: 'market-pattern'`. Don't pass a `namespace` arg — ReasoningBank routes it; on bridge unavailability the fallback writes to the reserved `pattern` namespace with `controller: 'memory-store-fallback'` (see ruflo-agentdb ADR-0001).
- **Plain store (namespace-routable)**: `mcp__claude-flow__memory_store --namespace market-patterns` — this DOES respect the `market-patterns` namespace because `memory_*` is namespace-routed.
6. **Report** -- display: pattern name, date range, direction, reliability, suggested action
## CLI alternative
```bash
npx @claude-flow/cli@latest memory search --query "bullish reversal patterns" --namespace market-patterns
npx @claude-flow/cli@latest memory store --key "pattern-AAPL-2026-05-04-doji" --value '{...}' --namespace market-patterns
```
---
name: market-pattern-4aee6964ffef
description: Detect and classify candlestick patterns from ingested OHLCV data
argument-hint: "<symbol> [--period 1D]"
allowed-tools: mcp__claude-flow__memory_search mcp__claude-flow__memory_list mcp__claude-flow__memory_store mcp__claude-flow__agentdb_pattern-store mcp__claude-flow__agentdb_pattern-search mcp__claude-flow__ruvllm_hnsw_route Bash
---
# Market Pattern
Scan ingested OHLCV data for known candlestick patterns, classify them by type and reliability, and store for future reference.
## When to use
When you need to identify candlestick patterns (doji, hammer, engulfing, head-shoulders, etc.) in market data. Requires data to be ingested first via `market-ingest`.
## Steps
1. **Load candles** -- call `mcp__claude-flow__memory_search` (or `memory_list`) on the `market-data` namespace to retrieve normalized OHLCV data for the symbol and period. The `memory_*` tool family routes by namespace; the `agentdb_hierarchical-*` family does NOT (it routes by tier), so use `memory_*` for namespaced reads.
2. **Scan for patterns** -- iterate through candle sequences looking for:
- Single-candle: doji (open ~= close), hammer (long lower wick), inverted hammer
- Two-candle: bullish/bearish engulfing
- Three-candle: morning star, evening star, three white soldiers, three black crows
- Multi-candle: head & shoulders, double top/bottom, cup & handle
3. **Classify** -- for each detection, assign: pattern name, type (reversal/continuation), direction (bullish/bearish), reliability score (0.0-1.0)
4. **Rank** -- sort by reliability score descending
5. **Store** -- two paths (per ruflo-cost-tracker ADR-0001 dual-path pattern):
- **Pattern-store (typed, recommended)**: `mcp__claude-flow__agentdb_pattern-store` with `type: 'market-pattern'`. Don't pass a `namespace` arg — ReasoningBank routes it; on bridge unavailability the fallback writes to the reserved `pattern` namespace with `controller: 'memory-store-fallback'` (see ruflo-agentdb ADR-0001).
- **Plain store (namespace-routable)**: `mcp__claude-flow__memory_store --namespace market-patterns` — this DOES respect the `market-patterns` namespace because `memory_*` is namespace-routed.
6. **Report** -- display: pattern name, date range, direction, reliability, suggested action
## CLI alternative
```bash
npx @claude-flow/cli@latest memory search --query "bullish reversal patterns" --namespace market-patterns
npx @claude-flow/cli@latest memory store --key "pattern-AAPL-2026-05-04-doji" --value '{...}' --namespace market-patterns
```
Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.