blitz-app-smoke

Pass

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill is configured to automatically install the 'blitz' package globally using npm install -g if the tool is not found on the system path.
  • [COMMAND_EXECUTION]: Extensive use of shell commands in run.sh to interact with the iOS development toolchain, including xcodebuild for listing and building projects, and xcrun simctl for simulator management.
  • [COMMAND_EXECUTION]: The run.sh script utilizes bash -lc for installing dependencies, which executes commands within a login shell environment.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting data from untrusted local files.
  • Ingestion points: Configuration values are read from codex.blitz.toml in the project root.
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded content are present in the parsing logic.
  • Capability inventory: Includes subprocess calls for building and testing code via xcodebuild, installing software via npm, and file system writes for artifacts in run.sh.
  • Sanitization: Configuration values are extracted via awk and used as arguments in shell commands; while they are quoted in the script, they lack explicit validation against a strict schema.
Audit Metadata
Risk Level
SAFE
Analyzed
May 15, 2026, 02:03 PM
Security Audit — agent-trust-hub — blitz-app-smoke