Security Audit

existential-birds/beagle:plugins/beagle-analysis/skills/llm-judge

github.com/existential-birds/beagle

AI SkillCommit aed6fce73b77

TRUSTED

Scanned 1 day ago

Critical

Immediate action required

High

Priority fixes suggested

Medium

Best practices review

Low

Acknowledged / Tracked

Trust Assessment

existential-birds/beagle:plugins/beagle-analysis/skills/llm-judge received a trust score of 85/100, placing it in the Mostly Trusted category. This skill has passed most security checks with only minor considerations noted.

SkillShield's automated analysis identified 1 finding: 0 critical, 1 high, 0 medium, and 0 low severity. Key findings include Indirect Prompt Injection via Untrusted Spec and Repository Content.

The analysis covered 4 layers: Manifest Analysis, Static Code Analysis, Dependency Graph, LLM Behavioral Safety. All layers scored 70 or above, reflecting consistent security practices.

Last analyzed on July 20, 2026 (commit aed6fce7). SkillShield performs automated 4-layer security analysis on AI skills and MCP servers.

Layer Breakdown

Manifest Analysis

100%

Static Code Analysis

100%

Dependency Graph

100%

LLM Behavioral Safety

85%

Behavioral Risk Signals

Filesystem Write

1 finding

Shell Execution

1 finding

Dynamic Code

1 finding

Excessive Permissions

1 finding

Security Findings1

Severity	Finding	Layer	Location
HIGH	Indirect Prompt Injection via Untrusted Spec and Repository Content The skill reads the contents of the spec file (`SPEC_CONTENT`) and the gathered facts from the repositories (`ALL_FACTS_JSON`) and interpolates them directly into the prompts for the Phase 1 and Phase 2 LLM agents without any delimiters, sanitization, or defensive framing. This makes the subagents highly vulnerable to indirect prompt injection. An attacker who controls the spec file or the repository being evaluated can inject malicious instructions to hijack the LLM, manipulate the evaluation results, or potentially execute unauthorized actions if the agent has tool access. Wrap untrusted content in clear, unique delimiters (e.g., <spec_document>...</spec_document>) and explicitly instruct the LLM to treat the enclosed content strictly as data, never as instructions. Implement validation or filtering to detect prompt injection patterns in the spec file and repository facts before passing them to the LLM.	LLM	SKILL.md:113

Scan History

Embed Code

[![SkillShield](https://skillshield.io/api/v1/badge/46130474ebd5c857.svg)](https://skillshield.io/report/46130474ebd5c857)