knowledge-graph — Claude Code skill: memory and setup
The knowledge-graph skill keeps durable facts about people, companies, and projects in a workspace graph. Use it to carry useful context between sessions and look it up without reopening long transcripts.
MIT ★ 239 updated 2026-08-30
npx skills add jdrhyne/agent-skills --skill knowledge-graph -a claude-codeWhat knowledge-graph does
The skill extracts lasting facts from recent notes and conversation history, then appends them to an entity’s facts.jsonl. Each entity also gets a concise current snapshot in summary.md; weekly synthesis refreshes that snapshot from active facts. Older entries stay in the record and are marked superseded when circumstances change.
The graph lives under life/areas/, with separate folders for people, companies, and projects. Create its base folders with mkdir -p life/areas/people life/areas/companies life/areas/projects. For lookups, the skill reads the summary first and opens the fact log only when details are requested or the summary is stale. It does not load the graph into every conversation, and it avoids storing secrets or highly personal data unless the user explicitly asks.
What it helps with
- preserve lasting details from recent conversations
- organize context by person, company, or project
- refresh entity summaries from current facts
- look up known context without reopening long transcripts
Example prompts for the agent
- Save the lasting details about Project Atlas from our recent conversation to the knowledge graph.
- Refresh the weekly summaries for entities that gained new facts.
- What does the graph know about Project Atlas? Include the underlying facts if I ask for detail.
How to install knowledge-graph
1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill knowledge-graph -a claude-code2Manually: put the files in place
~/.claude/skills/knowledge-graph/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/knowledge-graph/
1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill knowledge-graph -a codex2Manually: put the files in place
.agents/skills/knowledge-graph/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill knowledge-graph -a cursor2Manually: put the files in place
.cursor/skills/knowledge-graph/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install jdrhyne/agent-skills clawdbot/knowledge-graph/SKILL.md2Manually: put the files in place
.github/skills/knowledge-graph/SKILL.md1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill knowledge-graph -a gemini-cli2Platform CLI
gemini skills install https://github.com/jdrhyne/agent-skills --path clawdbot/knowledge-graphManually: put the files in place
.devin/skills/knowledge-graph/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about knowledge-graph
How do I install knowledge-graph in Claude Code?
Run npx skills add jdrhyne/agent-skills --skill knowledge-graph -a claude-code. The skill file is at ~/.claude/skills/knowledge-graph/SKILL.md.
What does the knowledge-graph skill do?
It records durable information about people, companies, and projects in an entity graph, then builds concise summaries from active facts.
How does knowledge-graph handle changed facts?
It appends new entries to facts.jsonl and marks outdated entries as superseded instead of deleting them. Each entity’s short snapshot is kept in summary.md.
Source text
SKILL.md · MIT
---
name: knowledge-graph
description: Three-Layer Memory System — automatic fact extraction, entity-based knowledge graph, and weekly synthesis. Manages life/areas/ entities with atomic facts and living summaries.
metadata: {"version":"1.1.0","openclaw":{"emoji":"🧠"}}
permissions:
- exec: "Uses local filesystem commands when creating entity folders or scheduled knowledge-graph jobs."
- file_write: "Appends facts, summaries, and daily-note synthesis inside the workspace knowledge graph."
---
# Knowledge Graph Skill
Maintain a lightweight, append-only entity graph that compounds durable facts across sessions.
## When to Use
- Extract durable facts from recent work or conversation history
- Rewrite entity summaries from active facts
- Answer "what do we know about X?" without reopening large transcripts
- Keep shared context for people, companies, and projects inside the workspace
## Data Model
Store the graph under:
```text
<workspace>/life/areas/
people/<slug>/
companies/<slug>/
projects/<slug>/
```
Each entity folder should contain:
- `summary.md` for the short, current snapshot
- `facts.jsonl` for atomic, append-only facts
Use one JSON object per line:
```json
{
"id": "<slug>-NNN",
"fact": "Plain-English fact",
"category": "relationship|milestone|status|preference|context|decision",
"ts": "YYYY-MM-DD",
"source": "conversation|manual|inference",
"status": "active|superseded",
"supersedes": "<older-id>"
}
```
## Fact Rules
- Keep facts atomic. One durable fact per entry.
- Append new facts instead of rewriting history.
- When something changes, add a new fact and mark the old one as superseded.
- Skip ephemera, greetings, speculation, and low-value chatter.
- Check existing facts before adding duplicates.
Durable facts usually include:
- role or relationship changes
- key decisions
- long-lived preferences
- major project milestones
- stable operating context
## Workflows
### Fact Extraction
1. Read the recent daily note and the recent conversation window.
2. Identify durable facts worth preserving.
3. Resolve entity type and slug.
4. Create the entity folder if it does not exist.
5. Append new facts to `facts.jsonl`.
6. Note extraction activity in the daily note if the workspace uses one.
### Weekly Synthesis
1. List entities changed during the week.
2. Load active facts only.
3. Rewrite `summary.md` in 3 to 8 concise lines.
4. Ensure contradicted facts are marked superseded.
5. Record a short synthesis note in the daily log if applicable.
### Entity Lookup
1. Read `summary.md` first.
2. Open `facts.jsonl` only if the summary is stale or the user asked for detail.
3. Fall back to broader memory search only when the entity is missing from the graph.
## Low-Token Recall
Recall should be triggered, not automatic.
- Recall when the user names a tracked person, company, or project.
- Recall when the user explicitly asks to remember, recall, or summarize prior context.
- Inject only the short summary by default.
- Avoid loading raw facts unless the user asked for specifics or contradictions need resolution.
## Setup
Create the core directories once:
```bash
mkdir -p life/areas/people life/areas/companies life/areas/projects
```
If multiple agents share one workspace, point them at the same `life/` directory so they operate on the same entity store.
## Safety Boundaries
- Do not store sensitive secrets, credentials, or highly personal data unless the user explicitly asked for it.
- Do not create entities or facts for casual chat that has no durable value.
- Do not inject the graph into every conversation by default.
- Do not delete historical facts; supersede them with a newer fact instead.
© jdrhyne · MIT · jdrhyne/agent-skills