JEP Primitive Skills

JEP Primitive Skills — Atomic Reference Implementations of Judge, Delegate, Terminate, Verify for Agent Collaboration Grammar

Security Scan

The skill's code, runtime instructions, and declared dependencies are coherent with its stated purpose (implementing four JEP primitives) and do not request unrelated credentials or perform unexpected I/O or network exfiltration.

Purpose & Capability

Name/description (J/D/T/V primitives) match the included Python implementation, FastAPI entrypoint, tests, and examples. Declared dependencies (fastapi, uvicorn, pydantic) are appropriate for an HTTP API exposing these primitives.

Instruction Scope

SKILL.md and manifest point to skill.api:app and define JSON schemas and endpoints; the code implements those endpoints and returns JEP-04-compliant events. The skill accepts an optional 'signature' field but does not perform cryptographic verification of signatures anywhere in the code — this is a functional omission (not necessarily malicious) but relevant to trust of issuer identity and should be addressed before relying on the skill in adversarial contexts.

Install Mechanism

No install spec provided (instruction-only), but packaged code includes a requirements.txt listing standard pip packages. No downloads from arbitrary URLs or archive extraction are present in the bundle.

Credentials

The skill does not require environment variables, credentials, or config paths. It neither reads secrets nor asks for unrelated tokens — requested resources are proportionate to an HTTP API providing event construction.

Persistence & Privilege

always is false and the skill does not request elevated privileges, modify other skills, or persist into agent configuration. It exposes an HTTP API when run (normal for a service) and can be invoked autonomously by the agent per platform defaults.

Assessment

This package appears internally consistent and implements the four primitives as described. Before installing or exposing it, consider: (1) provenance — the bundle has no homepage and a single author email; verify you trust the source. (2) Signature handling — the API accepts a 'signature' field but does not verify signatures; if you need authenticated events, add signature verification and key management. (3) Network exposure — running uvicorn will open an HTTP endpoint; avoid exposing it publicly without TLS, authentication, or a gateway. (4) Operational integration — the code produces JEP-04-compliant events but does not itself push events to downstream systems; ensure any downstream consumers enforce access control and validation. Finally, run the included tests and review the example before deploying in production.

Atomic Reference Implementations of Judge, Delegate, Terminate, Verify

The Four Primitives

Primitive	Verb	Purpose
Judge	J	Initiate an observation assertion
Delegate	D	Transfer authority to another agent
Terminate	T	Close the lifecycle of a prior assertion
Verify	V	Cross-validate an existing assertion

Complete Cognitive Interaction Algebra

These four primitives form a complete algebra:

• J creates state.
• V validates state.
• D distributes responsibility for state.
• T invalidates state.

Any complex multi-agent collaboration decomposes into ordered combinations of these four atoms. No fifth primitive is required.

Protocol Alignment

All events produced are strict JEP-04 compliant:

• jep: "1"
• verb: J/D/T/V
• who: issuer DID
• when: Unix timestamp
• what: SHA-256 multihash of payload
• nonce: UUIDv4
• aud: target
• ref: chain link
• sig: JWS signature
• task_based_on: JAC-01 parent task hash

Integration

Events feed directly into:

• Determinability-Checker — causal sufficiency verification
• COE-Consensus — shared world state formation
• JEP-Guard-Audit — compliance chain generation

---

Cognitive Emergence Lab
[e-mail]