Pulse — Capital Inflow Detector
Identifies assets accelerating in capital concentration before they become crowded positions. Uses publicly available HL market data as proxy signals for institutional flow detection.
Signal Types
| Signal | Trigger | Confidence |
|---|---|---|
| IMMEDIATE_MOVER | OI +15% AND volume 5x surge | 100 |
| VOLUME_SURGE | Recent 4h volume > 3x average | 70 |
| OI_BREAKOUT | OI jumps 8%+ above baseline | 60 |
| FUNDING_FLIP | Funding rate reversal or 50%+ acceleration | 50 |
Direction Classification
Majority vote across available signals:
- Funding rate sign -> directional bias
- Price breakout direction
- Volume surge + price momentum
Quality Filters
- Erratic detection (rank bouncing -> filtered)
- Minimum 24h volume ($500K default)
- Minimum scan history for baseline (2 scans)
Usage
bashhl pulse once # Single scan hl pulse run --tick 60 # Continuous (60s intervals) hl pulse once --json # JSON output hl pulse once --mock # Mock data hl pulse status # Last scan results hl pulse presets # List presets
Agent Mandate
You are the Pulse capital inflow detector. Your job is to catch capital inflow signals BEFORE the crowd. You detect timing — Radar detects setups. Together they form the APEX entry pipeline.
RULES:
- IMMEDIATE_MOVER is the only signal strong enough for standalone entry
- All other signals require Radar confirmation (score > 170)
- ALWAYS check baseline history — signals from assets with < 2 scans are unreliable
- NEVER act on erratic assets (rank bouncing between scans)
- Report direction with every signal — directionless signals are useless
Decision Rules
| Signal Type | Confidence | Standalone Entry? | Action |
|---|---|---|---|
| IMMEDIATE_MOVER | 100 | YES | Enter immediately — rare, high-conviction |
| VOLUME_SURGE | 70 | NO | Check Radar score. Enter if > 170 |
| OI_BREAKOUT | 60 | NO | Check Radar score. Enter if > 200 |
| FUNDING_FLIP | 50 | NO | Informational only — do not enter on funding alone |
| Multiple signals same asset | Varies | YES if combined > 150 | Compound conviction — enter with caution |
| Baseline State | Action |
|---|---|
| < 2 scans in history | Ignore all signals — insufficient baseline |
| Erratic rank changes | Ignore — likely data noise or manipulation |
| Consistent acceleration | High confidence — act on signal |
| Signal appeared then disappeared | Stale — do not enter |
Anti-Patterns
- Acting on FUNDING_FLIP alone: Funding flips are noisy and low-conviction. Only useful as confirming signal.
- Entering on VOLUME_SURGE without Radar: Volume surges occur on news, liquidations, and wash trading. 30% are false positives without Radar confirmation.
- Ignoring the erratic filter: Assets with bouncing rankings are being manipulated or have unstable liquidity. The erratic filter exists for a reason.
- Running too frequently: Pulse needs time between scans to build baselines. Running every 10s wastes API calls without improving signal quality. Use 60s minimum.
Error Recovery
| Error | Cause | Fix |
|---|---|---|
No signals detected | Quiet market | Normal — no action needed |
Baseline too short | First 2 scans | Wait for more scan history to accumulate |
API timeout on candle fetch | Rate limit | Increase tick interval or reduce parallel fetches |
Stale signal (> 5 min old) | Delayed processing | Re-scan before acting — signal may have decayed |
Composition
Pulse is a sub-component of APEX (runs every tick). Pairs with Radar — Radar identifies high-quality setups, Pulse detects the optimal entry timing via capital flow signals. When used standalone, always cross-check with hl radar once.
Cron Template
bash# Standalone Pulse scan every 60s during trading hours * 8-20 * * 1-5 cd ~/agent-cli && hl pulse once --json >> data/pulse/signals.jsonl 2>&1
