new-workspace

Pass

Audited by Gen Agent Trust Hub on May 7, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes standard system and development tools (mkdir, cp, git, gh) to perform file system operations and repository management.
  • [DATA_EXPOSURE]: The skill reads identity and environment facts from the local ~/.claude/CLAUDE.md file to personalize the new workspace. It also creates public GitHub repositories by default as stated in its documentation; users can override this with the --private or --local-only flags.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests data from local files (~/.claude/CLAUDE.md, config.json) and user arguments without explicit boundary markers or sanitization, though it primarily extracts structured facts for templating.
  • Ingestion points: ~/.claude/CLAUDE.md, config.json, $ARGUMENTS.
  • Boundary markers: Absent.
  • Capability inventory: mkdir, cp, git, gh.
  • Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
May 7, 2026, 09:14 PM
Security Audit — agent-trust-hub — new-workspace