skills-search

Warn

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is designed around the direct execution of shell commands using the Bash tool to manage system configurations and install software.
  • [EXTERNAL_DOWNLOADS]: The skill uses npx and the ccpm utility to fetch and install packages from an external registry. This involves downloading and executing remote code at runtime. While the bootstrap tool @daymade/ccpm is provided by the skill's author, the registry allows the installation of arbitrary third-party skills.
  • [INDIRECT_PROMPT_INJECTION]: The skill maps user-provided strings, such as search queries and skill names, directly into shell command arguments (e.g., ccpm search <query>). The instructions do not define any sanitization, validation, or boundary markers for this data, which could allow a malicious user to perform command injection by providing input designed to break out of the command string.
  • [PROMPT_INJECTION]: The instructions contain behavioral directives that explicitly tell the agent to bypass user confirmation for shell execution. Directives like "You MUST directly execute... Do NOT show the user a command and ask them to copy-paste it" and "Always execute directly" attempt to override standard safety protocols that require human-in-the-loop review for shell commands.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 8, 2026, 03:07 AM