claude-pm-skills

Pass

Audited by Gen Agent Trust Hub on Oct 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill functions as an installer for a suite of 15 product management skills hosted on a third-party GitHub repository (pratikshadake/claude-product-management-skills). It directs the agent to fetch content from this external source.
  • [COMMAND_EXECUTION]: The skill provides specific bash commands for the agent to execute, including npx skills add to install the external package. These commands use the -y flag to bypass confirmation and include options for global installation.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a management layer that installs other instructions. This presents an indirect injection surface where the agent's future behavior is governed by the external content it is instructed to download.
  • Ingestion points: The skill performs file existence checks at ~/.agents/skills/problem-clarity/SKILL.md and .agents/skills/problem-clarity/SKILL.md to determine installation status.
  • Boundary markers: No specific delimiters or safety warnings are provided for the content to be installed.
  • Capability inventory: The skill possesses the ability to execute shell commands and modify the local file system via the npx skills tool.
  • Sanitization: There is no validation or sanitization of the remote package content before installation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 22, 2026, 09:34 PM