matrix-github-project

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses tools like matrix run and send_terminal_input to execute commands on a remote VPS for repository probes and coding tasks.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted instructions from external repositories which creates a vulnerability surface for indirect prompt injection.
  • Ingestion points: SKILL.md (section: 'Understand and change the project') reads AGENTS.md, CLAUDE.md, and README files to determine coding steps.
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious embedded directives in the read files.
  • Capability inventory: The skill possesses broad execution capabilities, including matrix run, codex, and claude command execution across multiple files.
  • Sanitization: While the skill uses '--' to prevent shell command injection from user input, it lacks sanitization or filtering for natural language instructions parsed from the repository files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 06:10 PM
Security Audit — agent-trust-hub — matrix-github-project