article-agent-context-research
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION] (HIGH): The workflow in
SKILL.mdexplicitly calls for executingnpm run grok:context -- --topic \"...\". Thetopicparameter is interpolated directly from user input or content found indata/idea-creation/*.md, which can be manipulated to execute arbitrary commands. - [REMOTE_CODE_EXECUTION] (HIGH): The reliance on shell-based execution with untrusted data allows for command injection (e.g., using
;,|, or backticks) which could lead to full system compromise. - [PROMPT_INJECTION] (HIGH): Category 8 finding: The skill ingests untrusted data from local files (
data/idea-creation/) and user queries to drive high-privilege operations (shell execution). Evidence: 1. Ingestion point:data/idea-creation/*.mdand user 'Intake'. 2. Boundary markers: Absent. 3. Capability inventory:npm runsubprocess execution. 4. Sanitization: None detected. A malicious idea report could inject shell commands into thetopicvariable. - [DATA_EXFILTRATION] (MEDIUM): The skill is designed to read potentially sensitive research data from
data/article-research/*and send it to an external process (grok:context), which may leak internal information to third-party services.
Recommendations
- AI detected serious security threats
Audit Metadata