rlama

Fail

Audited by Snyk on Mar 2, 2026

Risk Level: CRITICAL
Full Analysis

CRITICAL E004: Prompt injection detected in skill instructions.

  • Potential prompt injection detected (high risk: 1.00). The prompt claims "Runs 100% locally" but then explicitly documents and instructs how to use external/cloud providers (OpenRouter, TogetherAI, OpenAI, custom endpoints and API keys), which contradicts and is deceptive relative to the skill's stated local-only purpose.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (high risk: 1.00). The skill explicitly ingests public website content via commands like "rlama crawl-rag", "rlama crawl-add-docs", and "rlama web-watch" (see SKILL.md and references/rlama-commands.md), which pulls arbitrary third-party webpages into the RAG that the agent will read and synthesize from, allowing untrusted web content to influence tool behavior.
Audit Metadata
Risk Level
CRITICAL
Analyzed
Mar 2, 2026, 09:44 PM