signal-synthesizer

compatibility	Python 3.8+, standard library only. No network. No third-party packages.
description	Combines per-agent DSE sub-scores (technical, fundamental, smart-money, sentiment, macro, volume-flow) into two synthesized signals — a long-term Investment signal and a short-term Momentum signal — each with a 1-10 DSE Composite Score, rating, confluence/conflict logic, and microstructure suppression. Use when the user wants to fuse multiple analyses into a final call, asks "what's the overall signal", "combine the agent scores", "should I invest or trade this", or needs the Mode Router / composite score for a Dhaka Stock Exchange ticker.
license	Apache-2.0
metadata	{"author":"stock-buddy","github-path":"skills/signal-synthesizer","github-pinned":"main","github-ref":"refs/heads/main","github-repo":"https://github.com/kuntal-r-d/my-skills","github-tree-sha":"0ffa31f0013f117beebca0ce0594b87e59889c89","mode":["investment","momentum"],"prd_refs":["PRD-002:REQ-022","PRD-002:REQ-111","PRD-002:REQ-097"],"version":"0.1.0"}
name	signal-synthesizer

Signal Synthesizer

When to use

Activate when individual committee skills (technical-analysis, fundamental-analysis, smart-money-flow, sentiment-news, macro-regime, …) have each produced a score and the user wants the final fused call: "combine these into one signal", "what's the overall rating for GP", "is this an investment or a trade", "give me the composite score". This skill is the Mode Router + Synthesizer (PRD-002 REQ-022) and emits the DSE Composite Score (REQ-111) for both modes at once.

What it does

Takes an agents map of sub-scores and produces two independent signals:

1. Investment — weighted toward fundamentals/smart-money/macro. Emits a rating, fair-value/entry-zone note, a thesis, and exit conditions.
2. Momentum — weighted toward technical/volume-flow. Fundamentals act as a veto (a strongly broken business caps a momentum trade at stand_aside).

It applies three governance rules on top of the weighted averages:

• Confluence — a high-conviction call (|weighted| ≥ 0.5) is only kept if ≥ 2 independent agents agree in sign with |score| ≥ 0.3; otherwise it is downgraded to a moderate rating.
• Strong conflict — if a strongly bullish agent (≥ 0.5) and a strongly bearish agent (≤ -0.5) coexist, the signal is stand_aside (not a diluted average), and the conflict is explained.
• Microstructure suppression (REQ-097) — if the stock is at a circuit limit (limit_up/limit_down), on a floor price, or halted, the Momentum signal is suppressed (price discovery is interrupted); the Investment signal is still emitted with a microstructure flag.

Each weighted score (-1..+1) maps to a DSE Composite Score 1-10 (rounded), and per-agent contributions (weight × score) are returned so the composite is fully decomposable.

How to run

python3 scripts/synthesize.py --input data.json --pretty
cat data.json | python3 scripts/synthesize.py

Input shape:

{
  "ticker": "GP",
  "as_of": "2026-06-19",
  "agents": {
    "technical":   {"score": 0.40, "confidence": 0.70},
    "fundamental": {"score": 0.60, "confidence": 0.80},
    "smart_money": {"score": 0.30, "confidence": 0.60},
    "sentiment":   {"score": 0.20, "confidence": 0.50},
    "macro":       {"score": -0.10, "confidence": 0.60},
    "volume_flow": {"score": 0.35, "confidence": 0.65}
  },
  "microstructure": {"circuit_state": "normal", "floor_price": null, "halted": false}
}

volume_flow is optional — if absent it is derived from the technical score. Scores are clamped to [-1, +1] and confidences to [0, 1].

Interpreting output

• investment and momentum blocks each carry score (-1..+1), composite_1_10, rating, confidence, and decomposable contributions.
• Ratings: strong_buy, buy, hold, sell, stand_aside, suppressed.
• stand_aside means the synthesizer found a genuine conflict — do not read it as a weak buy. suppressed means price discovery is interrupted (circuit/floor/halt).
• flags carry penalties and microstructure notes; they are never silently dropped.

Notes

Weights, the confluence/conflict logic, the composite-score mapping, and microstructure suppression are documented in references/SYNTHESIS.md. Output is educational analysis only, never financial advice.