linkedin-carousel-outline
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and follow instructions contained in local files (
.liftli/voice.mdor~/.liftli/voice.md). While this allows for personalization, it creates a surface where external data can influence the agent's behavior. - Ingestion points: Local voice configuration files (
.liftli/voice.md,~/.liftli/voice.md) and material gathered from chat. - Boundary markers: The skill lacks explicit delimiters to distinguish between system instructions and data ingested from the voice files.
- Capability inventory: Text generation for LinkedIn posts and recommendation of external tool invocations.
- Sanitization: No sanitization or filtering of the content within the voice files is specified.
- [EXTERNAL_DOWNLOADS]: The skill suggests downloading additional components and skills from the author's official namespace (
liftli-ai/linkedin-agent-skills). These resources are provided by the skill author and are part of the intended functionality. - [COMMAND_EXECUTION]: The documentation includes CLI commands for the user to manually add MCP servers (
claude mcp add) and install additional skills (npx skills add). These are standard integration steps and do not represent autonomous malicious execution.
Audit Metadata