Gift-Informed Community Health Agent — Design Spec

Date: 2026-04-04 Sub-project: B of 3

Overview

Extend the Community Health Agent with gift-economy community monitoring. Online communities are gift economies — members give attention, knowledge, support, and creative work. Healthy communities have high gift circulation. Unhealthy ones have stagnant pools or extractive patterns.

This builds on Sub-project A’s primitives (gratitude signals, generosity scoring, knowledge markers, circulation metrics) and the existing Community Health Agent prototype (Discord adapter, moderation + onboarding workflows, 3 agent templates).

What’s Being Added

1. Community Gift Flow Tracking

New workflow: Gift Flow Analysis — runs periodically to analyze community contribution patterns.

Nodes: - scan_recent_messages — pull recent Discord message signals, categorize by contribution type (question, answer, resource share, discussion, off-topic) - compute_gift_flow — for each member, compute: contributions given (messages that got replies/reactions), contributions received (replies to their messages), gift ratio (given/received) - detect_patterns — identify: top contributors, lurkers (consume but don’t contribute), dominators (high volume, low reciprocation), healthy circulators (balanced give/receive) - place_markers — place stigmergy markers for detected patterns: knowledge.pattern for community insights, quality.verified for healthy patterns, danger.untrusted for concerning patterns - generate_report — produce a community gift flow health report

2. New Discord Signal Types

Extend the Discord adapter’s signal mappings to emit richer signals that capture gift-economy dimensions:

• discord.help_given — when a message is a reply to a question (detected by context: original was a question, reply provides information)
• discord.resource_shared — when a message contains links, code blocks, or file attachments
• discord.gratitude_expressed — when a message contains thank-you patterns or gratitude reactions

These are derived signals — the adapter detects them from message_created events by analyzing content patterns.

3. Threshold Support for New Members

Extend the existing onboarding workflow with gift-economy awareness:

• After welcome_message, add a threshold_gift node that packages relevant community context for the new member (active topics, recent helpful threads, who to ask for what)
• After check_24h, if the member’s first message is a contribution (question or resource), emit a coord.gratitude signal amplifying their entry into the gift cycle

4. Community Generosity Scoring

Track per-member generosity using stigmergy markers:

• When a member helps another (reply to question, resource share), place a quality.high marker with their user ID in payload
• When a member only consumes (reads but never posts, or posts only self-promotional content), place a quality.low marker
• The Analyst agent aggregates these markers into a community-level gift health score

5. Hoarding Detection for Communities

New condition in the moderation workflow: detect members who dominate conversations (>30% of messages in a channel) without reciprocating (low reply-to-others ratio). This isn’t a ban-worthy offense — it’s a health signal that the community’s gift cycle is imbalanced.

• Emit stress.context_hoarding at the community level when detected
• Surface in the Loom dashboard via existing GiftEconomySummary component

Packages Touched

What’s NOT Changing

• No changes to the Stigmergy, Immune, Hypha, or Genesis packages — Sub-project A already provides the primitives
• No changes to the Discord adapter’s core architecture — just extending signal mappings
• No changes to the executor’s action types — gift flow is observational, not action-oriented
• The existing moderation and onboarding workflows remain unchanged — new workflow is additive

Testing Strategy

Extend the community health integration tests with: 1. Gift flow analysis workflow executes and produces markers 2. Threshold support sends context to new members 3. Help-given signal fires when reply pattern detected 4. Community-level hoarding detection fires on imbalanced participation