From gingko
Use for operator-grade Gingko memory operations the daily CLI does not expose. Triggers on "fork branch memory", "merge branch memory", "list memory branches", "prune branch", "set charter", "project charter", "force commit session", "flush memory now", "run maintenance", "decay memories", "consolidate memories", "refresh project summary", "get cluster". Only invoke when the user explicitly asks.
How this skill is triggered — by the user, by Claude, or both
Slash command
/gingko:gingko-operatorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The `gingko` CLI handles the daily recall + record loop. This skill covers operator-grade operations that live on MCP only: branch-isolated memory, clusters, the project charter, maintenance, and mid-workflow flush. **Invoke only when the user explicitly asks.**
The gingko CLI handles the daily recall + record loop. This skill covers operator-grade operations that live on MCP only: branch-isolated memory, clusters, the project charter, maintenance, and mid-workflow flush. Invoke only when the user explicitly asks.
Most MCP tools want project_id. Get it from the shell:
gingko memory project-id # e.g. "github.com--edlontech--gingko-mem"
gingko memory session-id # active session UUID, empty if none
Fork a feature branch's memory off main so unmerged scratchwork doesn't pollute the shared graph. List, fork, merge, inspect, prune.
mcp__plugin_gingko_gingko__list_branches
project_id: "<id>"
mcp__plugin_gingko_gingko__fork_branch
project_id: "<id>"
branch_name: "<git branch>"
from_branch: "main" # optional
mcp__plugin_gingko_gingko__get_merge_state
project_id: "<id>"
branch_name: "<source>"
into: "<target>"
mcp__plugin_gingko_gingko__merge_branch
project_id: "<id>"
branch_name: "<source branch>"
into: "main"
strategy: "fast-forward" | "three-way"
mcp__plugin_gingko_gingko__prune_branch_memory
project_id: "<id>"
branch_name: "<branch>"
The session-start hook auto-detects the current git branch when starting a session, so most users never need to call fork_branch by hand — only when reorganizing memory across branches.
Clusters are curated headline summaries — the canonical view of a topic. Recall is granular search; clusters are the table of contents. Slugs come from the cluster index printed by gingko memory session-primer.
mcp__plugin_gingko_gingko__get_cluster
project_id: "<id>"
slug: "auth-middleware"
The charter is the project's North Star — what it's for, what it isn't. Set it when the user articulates a durable mission, not on speculation.
mcp__plugin_gingko_gingko__set_charter
project_id: "<id>"
content: "<markdown>"
Before a multi-hour operation or a hard context switch within one session, force a commit so the work persists:
mcp__plugin_gingko_gingko__commit_session
session_id: "<current>"
The MCP call returns a fresh session_id — replace any cached value. The CLI equivalent closes the session entirely:
gingko memory close-session # commits and clears the active session pointer
After close-session, lifecycle hooks do not auto-rerun mid-session. To start a new session immediately, use mcp__plugin_gingko_gingko__start_session.
Background graph hygiene. Decay weakens stale links, consolidate merges near-duplicates, validate checks integrity.
mcp__plugin_gingko_gingko__run_maintenance
project_id: "<id>"
operation: "decay" | "consolidate" | "validate"
mcp__plugin_gingko_gingko__refresh_principal_memory
project_id: "<id>"
scope: "all" | "state" | "cluster"
refresh_principal_memory regenerates the summary regions in the primer on demand — normally it regenerates itself, but a forced refresh is useful after a large bulk import.
project_id — derived from the origin git remote. If multiple remotes or recently re-pointed, verify with git remote -v and gingko memory project-id.fork_branch on a branch with existing memory — operation fails; either pick a fresh name or prune_branch_memory first (destructive).npx claudepluginhub edlontech/edlon-agents-marketplace --plugin gingkoCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.