skills/rheinmir/setup/safe-change/Gen Agent Trust Hub

safe-change

Pass

Audited by Gen Agent Trust Hub on Jun 24, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use shell commands such as ls to find test files and a placeholder <test-cmd> to run tests. These are standard operations for a development-oriented agent skill and do not involve downloading or executing code from untrusted remote sources.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill involves reading and writing files within the project directory (e.g., modifying code and creating wiki drafts in llmwiki/). There are no instructions to access sensitive system files (like credentials or SSH keys) or to send data to external network domains.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions to generate wiki filenames and summaries. While this represents a potential surface for indirect injection, the instructions specify formatting rules (e.g., kebab-case, summary lengths) which act as natural constraints.
  • Ingestion points: User task description and module names are used to populate wiki templates.
  • Boundary markers: None explicitly defined.
  • Capability inventory: Shell execution (ls, test command) and file write operations.
  • Sanitization: None explicitly defined, though instructions require specific formatting for outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 24, 2026, 09:05 AM
Security Audit — agent-trust-hub — safe-change