BalajiSlides
Warn
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses imperative language such as 'MANDATORY' and 'REQUIRED BEFORE ANY ACTION' to force the agent to execute a specific background command before processing any user requests. This overrides the agent's default task prioritization.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command using
curlto a local endpoint (http://localhost:8888/notify). This unnecessary side-effect for a slide-design tool increases the security risk by requiring shell access for background network operations. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content to format slides, creating a surface for injection. 1. Ingestion points: User-provided topic, thesis, key claims, and evidence descriptions (SKILL.md, Workflows/CreateBalajiSlides.md). 2. Boundary markers: Absent; user input is interpolated directly into slide templates. 3. Capability inventory: Execution of
bashcommands viacurl(Workflows/CreateBalajiSlides.md). 4. Sanitization: Absent; no validation or escaping of user input is specified.
Audit Metadata