Edge Signal Aggregator

Aggregate and rank signals from multiple edge-finding skills (edge-candidate-agent, theme-detector, sector-analyst, institutional-flow-tracker) into a prioritized conviction dashboard with weighted scoring, deduplication, and contradiction detection.

No API

Download Skill Package (.skill) View Source on GitHub

Table of Contents

1. Overview

Combine outputs from multiple upstream edge-finding skills into a single weighted conviction dashboard. This skill applies configurable signal weights, deduplicates overlapping themes, flags contradictions between skills, and ranks composite edge ideas by aggregate confidence score. The result is a prioritized edge shortlist with provenance links to each contributing skill.

2. When to Use

  • After running multiple edge-finding skills and wanting a unified view
  • When consolidating signals from edge-candidate-agent, theme-detector, sector-analyst, and institutional-flow-tracker
  • Before making portfolio allocation decisions based on multiple signal sources
  • To identify contradictions between different analysis approaches
  • When prioritizing which edge ideas deserve deeper research

3. Prerequisites

  • Python 3.9+
  • No API keys required (processes local JSON/YAML files from other skills)
  • Dependencies: pyyaml (standard in most environments)

4. Quick Start 

python3 skills/edge-signal-aggregator/scripts/aggregate_signals.py \
  --edge-candidates reports/edge_candidate_agent_*.json \
  --edge-concepts reports/edge_concepts_*.yaml \
  --themes reports/theme_detector_*.json \
  --sectors reports/sector_analyst_*.json \
  --institutional reports/institutional_flow_*.json \
  --hints reports/edge_hints_*.yaml \
  --output-dir reports/

5. Workflow

Step 1: Gather Upstream Skill Outputs

Collect output files from the upstream skills you want to aggregate:

  • reports/edge_candidate_*.json from edge-candidate-agent
  • reports/edge_concepts_*.yaml from edge-concept-synthesizer
  • reports/theme_detector_*.json from theme-detector
  • reports/sector_analyst_*.json from sector-analyst
  • reports/institutional_flow_*.json from institutional-flow-tracker
  • reports/edge_hints_*.yaml from edge-hint-extractor

Step 2: Run Signal Aggregation

Execute the aggregator script with paths to upstream outputs:

python3 skills/edge-signal-aggregator/scripts/aggregate_signals.py \
  --edge-candidates reports/edge_candidate_agent_*.json \
  --edge-concepts reports/edge_concepts_*.yaml \
  --themes reports/theme_detector_*.json \
  --sectors reports/sector_analyst_*.json \
  --institutional reports/institutional_flow_*.json \
  --hints reports/edge_hints_*.yaml \
  --output-dir reports/

Optional: Use a custom weights configuration:

python3 skills/edge-signal-aggregator/scripts/aggregate_signals.py \
  --edge-candidates reports/edge_candidate_agent_*.json \
  --weights-config skills/edge-signal-aggregator/assets/custom_weights.yaml \
  --output-dir reports/

Step 3: Review Aggregated Dashboard

Open the generated report to review:

  1. Ranked Edge Ideas - Sorted by composite conviction score
  2. Signal Provenance - Which skills contributed to each idea (with evidence_id per contributor)
  3. Contradictions - Conflicting signals flagged for manual review
  4. Deduplication Log - Merged overlapping themes (verbatim re-ingested duplicates get no bonus)
  5. Invalid Signals - Scores that are NaN/infinite, negative, or out of declared range are dropped and reported, never converted into confidence

Step 4: Act on High-Conviction Signals

Filter the shortlist by minimum conviction threshold:

python3 skills/edge-signal-aggregator/scripts/aggregate_signals.py \
  --edge-candidates reports/edge_candidate_agent_*.json \
  --min-conviction 0.7 \
  --output-dir reports/

6. Resources

References:

  • skills/edge-signal-aggregator/references/signal-weighting-framework.md

Scripts:

  • skills/edge-signal-aggregator/scripts/aggregate_signals.py

7. Producer Contract and Scoring Rules (schema 1.1)

Since schema 1.1, aggregation is idempotent and strict about score validity (Issue #510):

  • evidence_id (optional, per item): stable evidence identity. Re-ingesting the same evidence (same skill + evidence_id) never adds a bonus. Without it a deterministic fallback is derived from skill, signal reference, title, tickers, and direction (timestamps excluded, so refreshed timestamps cannot forge a new identity). When copies differ, the newest observation wins: the latest parsed timezone-aware timestamp (date-only and naive values read as UTC) supplies the retained score. Equal, missing, or unparsable timestamps keep the highest raw score, then a stable key, so ingestion order cannot change the result. Only ISO-8601 strings (or YYYY-MM-DD) are ordered; epoch numbers and compact forms such as 20261010 count as missing timestamps. A newer, lower score replaces an older, higher one, and discarded copies are listed in deduplication_log[].discarded_copies.
  • score_scale (optional, per item or per document: "0-1" / "0-100"): declared scales are enforced strictly. NaN/infinity, negative, and out-of-range values are rejected into invalid_signals; a present-but-unparsable declaration ("", "foo") is rejected as missing_scale. When the key is absent, values in (1, 100] are assumed 0-100 and flagged (score_scale_assumed: true + scale_assumptions_report with the original value and the normalized_value); values > 100 are rejected.
  • Agreement rule: bonuses fire on agreement_tier = min(distinct evidence ids, distinct event fingerprints, distinct skills). Verbatim cross-skill restatements (identical normalized title + tickers) count once and cannot self-confirm; re-ingested duplicates earn nothing. Known limitation: event fingerprints are vocabulary-based, so paraphrased cross-skill restatements with disjoint word sets still earn the agreement bonus.
  • merge_bonus_per_duplicate (in custom weights YAML) is deprecated and ignored.