skills/ruvnet/ruview/Pair Programming/Gen Agent Trust Hub

Pair Programming

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines a large set of instructions and examples where the agent is expected to interact with the system via the claude-flow CLI tool. This includes starting and ending sessions, managing configurations, and performing Git operations like committing and stashing.
  • [INDIRECT_PROMPT_INJECTION]: The skill functions by analyzing and modifying user-provided source code, reading Git histories, and processing outputs from testing frameworks and linters. This creates an attack surface where malicious instructions hidden in the codebase could influence the agent's behavior.
  • Ingestion points: Project source files, Git diffs, test logs, and linting reports.
  • Boundary markers: None are explicitly defined in the skill's instructions to separate user data from agent instructions.
  • Capability inventory: The agent has capabilities to write files (/implement, /refactor), execute shell commands (via the claude-flow CLI and testing commands), and perform Git repository actions.
  • Sanitization: There are no documented mechanisms for sanitizing or escaping content retrieved from the analyzed files before it is processed by the agent.
  • [EXTERNAL_DOWNLOADS]: The skill's documentation recommends the installation of the claude-flow command-line utility from a public registry (npm install -g claude-flow@alpha).
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 07:48 PM