signal-factory

Classifies inbound signals, scores trust and risk, and emits fixed decision reports with receipts.

Updated Jun 20, 2026
One-click install
npx skills add https://github.com/Noetfield-Systems/SourceA --skill signal-factory-noetfield-systems
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: signal-factory
Source: https://github.com/Noetfield-Systems/SourceA/tree/main/.cursor/skills/signal-factory
Command: npx skills add https://github.com/Noetfield-Systems/SourceA --skill signal-factory-noetfield-systems

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It standardizes the triage of inbound commercial, partner, legal, support, and governance signals by enforcing a fixed classify-score-decide-receipt pipeline, so every signal gets a consistent decision report instead of ad-hoc judgment. ## Core Features & Use Cases - Fixed decision reports: Every run emits signal_summary, classification, trust/risk/automation/commercial scores, decision, next_action, receipt JSON, and a memory line in a strict order. - Hard risk override: Any signal with risk_score >= 4 is forced to route to human/legal review, and optional automation or commercial sections are suppressed. - Sender-claim tagging and entity hygiene: Sender assertions are tagged sender_declared and never restated as facts, while entity boundaries across Noetfield, TrustField, SourceA, WitnessBC, SG, and NOOS are preserved. - Use Case: A pricing demo request arrives for TrustField; run the core script to classify it as commercial_inquiry, score it, receive a decision with a receipt JSON, and verify the report against the schema. ## Quick Start Ask the assistant to analyze the text of an inbound signal with the signal factory and return the fixed decision report with its receipt JSON.

Frequently Asked Questions about signal-factory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I analyze an inbound signal with the signal factory?

Run the core script with the signal text, for example python3 scripts/signal_factory_core_v1.py --text "your signal text" --json. It returns the fixed decision report including classification, scores, decision, next_action, receipt, and memory line.

What happens when a signal has a high risk score?

A hard override applies: if risk_score is 4 or higher, the decision is forced to route and the next_action must send the signal to human or legal review. Optional automation_recipe and commercial_idea sections must not appear.

How are sender claims handled in decision reports?

All sender assertions are recorded with the tag sender_declared and source sender. They are never restated as verified facts in the signal_summary, implied_need, memory_line, or service patterns.

When do automation_recipe and commercial_idea sections appear?

They are gated strictly by the decision enum: automation_recipe appears only when decision is build_automation, and commercial_idea only when decision is create_service_pattern. No score threshold or fuzzy trigger applies.

Does the signal factory connect to Gmail, LinkedIn, or website forms?

No. Version 1 core explicitly excludes Gmail, LinkedIn, website form, and UI connections, and adapter hooks are empty null slots. The receipt field production_connected must remain false.

How do I verify a signal factory decision report?

Use python3 scripts/signal_factory_core_v1.py --verify-json path/to/report.json --json to validate a report, or run the six synthetic tests with --run-tests. The shell verifier scripts check the skill package structure and schema.