model-attack-paths

name	model-attack-paths
description	Build and rank paths from observed entry points through identity, cloud, code, and data relationships to crown jewels. Use when a Threat and Risk workflow needs the model stage.

Model Attack Paths

Contract

• Role: model.
• Accept: threat/intelligence/corpus/v1.
• Produce: threat/attack/graph/v1.
• Operate only through tools and integrations granted by TPSE core. A skill cannot expand its own authority.

Procedure

1. Confirm the organization, programme, scope, task, run, and source revision.
2. Read only the declared inputs and bounded AgentForum brief relevant to this work.
3. Build typed relationships from cited inputs. Do not create an edge or attribution without a reproducible basis.
4. Cite source artifact IDs and hashes for every material conclusion.
5. Emit the declared artifact contract, including confidence, freshness, provenance, and explicit gaps.

Integrity rules

• Never turn missing or failed collection into a passing or zero result.
• Never place credentials, secret values, private URLs, or unredacted sensitive samples in artifacts or AgentForum.
• Never mutate a customer or public system unless the selected tool declares that effect and core records the required approval.
• Prefer stable IDs and deltas so later runs can supersede rather than duplicate prior findings.