Skip to content
Back to skills

Graph

ASecurity

Record findings as a typed graph (claims, sources, decisions, with supports, cites, contradicts and depends_on edges) that parallel agents share, and verify with a deterministic grounding check that every claim traces to a source. Use for multi-agent research, deliverables whose claims must be traceable, or /strata:graph. For a linked knowledge base of pages, use /strata:wiki.

  • 5 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 11, 2026
ai-agentspythonnode

Security analysis

A100/100

Scanned November 11, 2026

npx -y skills add ahmadhammoud56/MOAA --skill graph --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Graph?

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

Security grade badge for Graph
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/ahmadhammoud56-graph/badge)](https://www.skillsdirectory.com/skills/ahmadhammoud56-graph)

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
description: Record findings as a typed graph (claims, sources, decisions, with supports, cites, contradicts and depends_on edges) that parallel agents share, and verify with a deterministic grounding check that every claim traces to a source. Use for multi-agent research, deliverables whose claims must be traceable, or /strata:graph. For a linked knowledge base of pages, use /strata:wiki.
argument-hint: "add | check | export [DIR]"
---

# Strata graph: findings with their sources

Request: `$ARGUMENTS`. `G` means `python "${CLAUDE_PLUGIN_ROOT}/graph.py"` (`py` on Windows), run as one standalone
command. Default folder: `kb/` (shared with `/strata:wiki`; `G init DIR` creates it).

"Graph engineering": once several agents work in parallel, findings must live in shared, typed state instead of
transcripts, and a script, not a model's impression, decides what is grounded. Each agent stays a loop; the graph is
the wiring and the evidence between them.

## Record
- Source: `G node DIR s-<slug> source "<title>" --ref <path or URL>` (a ref is required for grounding).
- Claim: `G node DIR c-<slug> claim "<one checkable statement>"`, then `G edge DIR c-<slug> supports s-<slug>`.
  A fact that holds only for a time gets `--valid-from YYYY-MM-DD` and, once replaced, `--valid-to YYYY-MM-DD`
  (write the node again); `G check` lists claims past their `valid_to` as expired and leaves them out of the score.
- Other types: `entity`, `decision`, `task`. Other edges: `cites`, `contradicts` (never overwrite a disagreeing claim;
  link it), `depends_on`, `derived_from`, `about`.
- Writing a node again with the same id replaces it (latest line wins). Each command appends one line, so parallel
  writers do not clash.

## Verify
`G check DIR` prints grounded/all claims (the score to drive to 1.0) and one specific failure per problem: a claim
with no source edge, a source with no ref, an edge to a missing node. It lists contradictions for a human or the
council to settle. Exit code 1 while anything fails, so it can gate a loop or a review.

## Use with agents
Put the folder path and the two record commands in each worker's brief (council workers, War Room workers, allies).
Before integrating results run `G check DIR`; send ungrounded claims back, or to the council's claim-table evidence
review, instead of accepting them. `G export DIR` writes `wiki/graph.md`, a Mermaid diagram of the graph.

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.

View all in ai-agents →

Comments

Loading comments…

Newsletter

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