jalopy1

DKG Working Memory Bridge — Agent Skill

Requires a running DKG v10 node. Auth token from `~/.dkg/auth.token` or `DKG_AUTH_TOKEN`.

jalopy1 0 Updated 1mo ago

Resources

15
GitHub

Install

npx skillscat add jalopy1/dkg-wm-bridge

Install via the SkillsCat registry.

SKILL.md

DKG Working Memory Bridge — Agent Skill

Deposit workspace artifacts into DKG v10 Working Memory with provenance and status tags.
Works with OpenClaw, Hermes, and any agent that produces text artifacts.

Setup

# First time: check connectivity and initialize
wm-bridge check
wm-bridge init --agent <name> --framework <openclaw|hermes|generic>

Config is saved to ~/.dkg/wm-bridge.json. Auto-detects peer ID from the node.

Commands

wm-bridge check

Verify DKG node connectivity and show agent identity.

wm-bridge init --agent <name> --framework <framework>

Create context graph, auto-detect peer ID, save config. Run once per agent.

wm-bridge ingest <file|directory>

Ingest markdown files into Working Memory. Each file becomes a named assertion with structured RDF.

Options: --recursive -r, --status <tag>, --tags <t1,t2>, --context-graph <id> -c, --dry-run

wm-bridge ingest-text <title> <text>

Ingest inline text (not from a file) into Working Memory.

wm-bridge promote <assertion-name>

Promote an assertion from Working Memory to Shared Memory.

wm-bridge discard <assertion-name>

Discard an assertion from Working Memory without promoting.

wm-bridge status

List artifacts currently in Working Memory with kind, status, and date.

wm-bridge info <assertion-name>

Show assertion details and lifecycle history.

When to Use

  • After writing daily memory notes — ingest them for durability and discoverability
  • After completing research — ingest findings so other agents can discover them
  • When curating long-term memory — ingest updates as knowledge artifacts
  • When collaborating — promote artifacts to Shared Memory for other agents

Multi-Agent Collaboration

Multiple agents can share a Context Graph:

# Agent 1 (OpenClaw)
wm-bridge init --agent MyAgent --framework openclaw -c team-research

# Agent 2 (Hermes)
wm-bridge init --agent Hermes --framework hermes -c team-research

# Both agents' promoted artifacts are visible to each other in Shared Memory

Trust Gradient

Artifacts flow through the v10 trust gradient:

  • Working Memory (private, per-agent) → draft / reviewed / promote-ready
  • Shared Memory (team-visible, gossiped) → promoted
  • Verified Memory (on-chain, Round 2) → verified-ready

Environment

Requires a running DKG v10 node. Auth token from ~/.dkg/auth.token or DKG_AUTH_TOKEN.