linkedin-reader

Warn

Audited by Gen Agent Trust Hub on Oct 20, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_CONTEXT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions direct the installation of a third-party CLI tool (@jackwener/opencli) and a Chrome extension from an unverified GitHub repository (jackwener/opencli). This introduces a dependency on external code that has broad access to the user's browser sessions.
  • [COMMAND_EXECUTION]: The skill operates by executing shell commands using the opencli utility to read LinkedIn timelines and search for job listings.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from LinkedIn, which could contain malicious instructions designed to manipulate the AI agent's logic or output.
  • Ingestion points: LinkedIn home feed posts and job search result descriptions retrieved in SKILL.md and references/commands.md.
  • Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when the agent processes the external LinkedIn data.
  • Capability inventory: Terminal output and text processing of social media data which the agent uses for financial research.
  • Sanitization: The skill does not implement any sanitization, filtering, or escaping of the data fetched from LinkedIn.
  • [DYNAMIC_CONTEXT_INJECTION]: SKILL.md utilizes the dynamic context pattern (!command) to execute a status check (opencli doctor) when the skill is loaded by the agent. While the command used here is for environment verification, it represents an automated execution at load time.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 20, 2026, 06:46 AM