eu-sparql-search

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The skill includes explicit, out-of-scope instructions to bypass platform permission checks (use bash_tool to "bypass all web_fetch permission restrictions") and to disable SSL verification, which are hidden/deceptive security‑bypass directives not required for describing the SPARQL API.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The skill contains explicit instructions to bypass platform network restrictions (use of bash_tool to evade web_fetch) and to disable SSL verification, which are deliberate actions that enable evasion of security controls and unsafe network access.

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

• Third-party content exposure detected (high risk: 0.85). The required workflow executes SPARQL via bash_tool against a public endpoint and then (optionally) fetches “full document text” from Cellar using curl, meaning outsider-authored web content is ingested as readable HTML/text into the agent’s LLM context.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill explicitly instructs the agent to "bypass all web_fetch permission restrictions" by using bash_tool and shows how to disable SSL verification and use curl to fetch URLs, which directs the agent to circumvent platform/network security controls.