Skip to content
Back to skills

Graph Add

ASecurity

Use when adding a new entity or fact to the knowledge graph.

  • 35 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 19, 2026
ai-agentsgobashnode

Security analysis

A100/100

Scanned September 19, 2026

npx -y skills add xcota/pos --skill graph-add --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Graph Add?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Graph Add
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/xcota-graph-add/badge)](https://www.skillsdirectory.com/skills/xcota-graph-add)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: graph-add
description: "Use when adding a new entity or fact to the knowledge graph."
version: 1.0
user_invocable: true
---
# /graph-add — Add Knowledge to the Graph

Quickly add a new entity or fact to the knowledge graph.

## Usage
`/graph-add` — interactive mode, asks questions
`/graph-add {type} {name} {facts}` — direct mode

## Entity Types
- `person` → `knowledge/people/{name}.md`
- `project` → `knowledge/projects/{name}.md`
- `concept` → `knowledge/concepts/{name}.md`
- `tool` → `knowledge/tools/{name}.md`
- `decision` → `knowledge/decisions/{date}-{name}.md`
- `research` → `knowledge/research/{date}-{name}.md`
- `business` → `knowledge/business/{name}.md`

## Steps

1. Determine type (ask if not provided).
2. Determine name (ask if not provided).
3. Collect key facts (ask if not provided).
3b. **Conflict check** — `grep -rl "{normalized subject}" knowledge/`. If an existing note asserts a DIFFERENT value for the same subject → do NOT overwrite: use conflict-as-record (see `knowledge/conflicts/AGENTS.md`) — create `knowledge/conflicts/<slug>-conflict-{date}.md` (both claims + provenance + `promotion_blocked: true`), downgrade the original to `disputed`, then stop and report to the user.
4. Identify related entities — search existing files:
   ```bash
   grep -rl "{related term}" knowledge/
   ```
5. Create the entity file:
   ```markdown
   ---
   type: {type}
   tags: [{relevant tags}]
   updated: {today}
   # epistemic (optional, enums: knowledge/conflicts/AGENTS.md): modality · epistemic_status (fresh single-source → extracted_candidate) · confidence · last_verified · conflicts: []
   ---
   # {Name}

   {Key facts as prose or bullet points}

   ## Connections
   - Related to [[{entity-1}]] — {why}
   - Related to [[{entity-2}]] — {why}
   ```
6. Update related entity files to add a backlink `[[{new-entity}]]`.
7. Report what was created and linked.

## File Naming
- Lowercase, hyphens for spaces: "Acme Corp" → `acme-corp.md`.
- Decisions and research are date-prefixed: `2026-01-05-chose-obsidian.md`.
- Max 60 chars.

## Rules
- Always add `[[wikilinks]]` to related entities.
- **MANDATORY: update backlinks in ≥1 related file** — a new entity MUST have ≥1 incoming link. If you create a file and no other file links to it, it's invisible in the graph. This is the #1 graph quality rule.
- Don't create duplicates — search first: `grep -rl "term" knowledge/`.
- Keep entity files focused: one entity = one file.
- Minimum content: YAML frontmatter + 1 paragraph + ≥2 outgoing `[[wikilinks]]` + ≥1 incoming backlink added to another file.

## Validation (run after every graph-add)
```bash
# Verify the new entity has incoming links
grep -rl "\[\[{new-entity-name}\]\]" knowledge/ | grep -v "{new-entity-name}.md"
# Must return ≥1 file. If 0 → FIX before committing.
```

## Anti-Pattern: Floating Nodes
NEVER create an entity without ensuring at least one other entity links to it.
Root cause when this slips: entities get created with outgoing links but nobody updates the hub files to link back. Always close the loop.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.