install-rules

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill manages local project configuration files within the .claude/rules/ directory, which is a standard practice for AI agent environments.
  • [COMMAND_EXECUTION]: Usage of the Bash tool is restricted to specific, non-destructive subcommands (git, ls, mkdir, diff) required for version tracking and directory management.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes rules from a plugin directory (Ingestion point) and installs them into the project using Precedence headers (Boundaries). Capabilities include filesystem Write and restricted Bash operations (Inventory). Risk is mitigated by hash-based verification in the vendor-owned .sd0x/install-state.json manifest and mandatory human confirmation via AskUserQuestion for conflicts (Sanitization).
  • [DATA_EXPOSURE]: No unauthorized data exposure detected. Tool use is limited to developer-centric Bash subcommands for local repository management with no evidence of sensitive data harvesting.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 01:44 PM
Security Audit — agent-trust-hub — install-rules