Back to skill

Security audit

jep-primitive-skills

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to be a disclosed local tool for creating unsigned draft events, with no evidence of hidden execution, persistence, or data exfiltration.

Install only in an environment where you are comfortable running a local FastAPI service and managing bearer tokens. Keep it bound to loopback unless you add TLS, rate limiting, and proxy controls; protect JEP_CLIENTS and trusted key configuration; and do not treat returned drafts as signed, verified, or action-executing events.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (8)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description suggests a narrowly scoped skill whose purpose is to construct explicitly unsigned J/D/T/V event drafts. The supplied code chunk is instead an executable example script that orchestrates a seven-step multi-agent workflow using primitive skills, generates multiple events of different types, links them causally, and prints detailed status/output. While it does involve creating J/D/T/V events, its primary purpose is demonstrative workflow execution, not a simple draft-construction utility. No dangerous external resource access is evident, but the actual behavior is materially broader and different from the declared purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared purpose says this skill constructs explicitly unsigned J/D/T/V event drafts, but the supplied code chunk contains no logic related to event creation, drafting, signing state, triggers, or permissions. It only defines metadata and a version string. Based on the provided code alone, the actual behavior does not match the declared purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The description says the skill constructs explicitly unsigned J/D/T/V event drafts. The code does include construction of a JEP04Event from a friendlier structure, but its behavior is broader and materially different in several ways. First, it is a full mapping layer with both to_jep04 and from_jep04, so it also parses/deserializes events, which is undeclared. Second, the output is not explicitly unsigned: sig is set to friendly.signature or an empty string, meaning signed input is preserved rather than rejected or stripped. Third, the adapter performs significant validation and enrichment—UUID canonicalization/version checks, timestamp conversion, reference selection, what-hash computation, reserved-extension enforcement, and creation of JAC/JEP extensions. That exceeds a narrow description of merely constructing unsigned event drafts.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description says the skill constructs explicitly unsigned event drafts for certain event types, implying a drafting/formatting role without security-sensitive behavior. The code instead provides broad security infrastructure: it canonicalizes JSON, signs events, verifies signatures against trusted keys, authenticates API callers via bearer tokens, enforces permissions and issuer scoping, and adds HTTP middleware and exception handlers. These are substantial undeclared capabilities and represent a materially different primary purpose from drafting unsigned events.

Content

No source excerpt is available for this finding.

Possible Typosquatting: 'uvicorn' resembles popular package 'gunicorn'

High
Category
Supply Chain
Confidence
70% confidence
Finding

Package name closely resembles a popular package, suggesting possible typosquatting. Attackers publish malicious packages with similar names to trick developers into installing them.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding

The skill metadata declares no explicit tool scope or permissions, yet the package is described as depending on environment-based authentication/configuration. Missing scope declarations make it harder to enforce least privilege and allow an agent or reviewer to underestimate what the skill may access at runtime. In a security-sensitive event-construction skill, this ambiguity increases the chance of unintended secret exposure or overbroad execution context.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The module docstring says this is a mapping layer that converts between FriendlyEvent and strict JEP04Event, implying semantic preservation. However, from_jep04 always sets assertion=None and derives verify_of from event.ref while also using extensions for other fields, so the reverse mapping does not actually reconstruct the original FriendlyEvent content.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

This code reads security-sensitive configuration from environment variables, including trusted keys and client authentication data via JEP_TRUSTED_KEYS and JEP_CLIENTS. While this is likely intentional for a security module, the file provides no user-facing warning, logging, or explanatory comment around that sensitive credential/configuration access beyond a general module docstring.

Content

No source excerpt is available for this finding.