jobs-evaluate

Fail

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions to download and execute an installation script from the author's own GitHub repository (solid-company/solid-jobs-skills). This involves piping remote content to bash or PowerShell to install the sjctl CLI tool.
  • [PROMPT_INJECTION]: The skill processes untrusted third-party data from job listings which could contain hidden instructions to manipulate the agent's evaluation results. Ingestion points: Job offer title, description, and company fields (SKILL.md). Boundary markers: No explicit delimiters are used to wrap the untrusted job content. Capability inventory: The skill uses the sjctl tool to read local profiles and save evaluation data (SKILL.md). Sanitization: The skill relies on natural language instructions warning the agent to ignore embedded commands rather than using structural sanitization or delimiters.
  • [COMMAND_EXECUTION]: The skill executes the sjctl CLI tool to interact with job offers, user profiles, and market data.
  • [EXTERNAL_DOWNLOADS]: The skill uses curl and irm to fetch installation scripts from a remote repository.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/solid-company/solid-jobs-skills/v0.6.0/scripts/install-sjctl.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Aug 14, 2026, 08:13 AM
Security Audit — agent-trust-hub — jobs-evaluate