skillshare

Fail

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The runner script scripts/run.sh facilitates the download and execution of a compiled binary from an untrusted GitHub repository (runkids/skillshare). This allows arbitrary code execution on the host machine.
  • [EXTERNAL_DOWNLOADS]: The skill documentation and runner script promote the 'curl | sh' pattern, a high-risk method for executing remote code without verification.
  • [EXTERNAL_DOWNLOADS]: The install command enables fetching and linking content from arbitrary, unverified third-party Git repositories and URLs, creating a large attack surface for untrusted code injection.
  • [COMMAND_EXECUTION]: The script scripts/run.sh uses shell commands (curl, tar, chmod, exec) to download, prepare, and run software directly from the internet.
  • [PROMPT_INJECTION]: The skill exposes AI tools to indirect prompt injection by syncing unverified remote content into their active skill directories.
  • Ingestion points: Remote Git repositories via references/install.md.
  • Boundary markers: None specified for the content of the linked skills.
  • Capability inventory: Modifies sensitive AI tool configuration directories (e.g., ~/.claude/skills).
  • Sanitization: Includes a security auditor mentioned in references/audit.md, though its effectiveness is unverified against the core runner script.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/runkids/skillshare/main/skills/skillshare/scripts/run.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Mar 1, 2026, 09:42 AM