flutter-sdk-changelog

Fail

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/build_flutter_sdk_changelog.py executes shell commands using subprocess.run with shell=True. The command string is formed by concatenating variables representing the Flutter SDK root path and git revisions. Since these variables can be influenced by environment settings or user-provided arguments and are not sanitized for shell metacharacters, this presents a potential command injection vulnerability.
  • [COMMAND_EXECUTION]: Instructions in SKILL.md suggest a command that pipes remote JSON data into a Python one-liner (python3 -c). While the execution logic is provided locally in the command string rather than fetched from the network, the practice of piping external data directly into a language interpreter is a sensitive pattern.
  • [EXTERNAL_DOWNLOADS]: The skill is configured to fetch the official Flutter release index from a Google Cloud Storage bucket and to access the Flutter source code repository on GitHub. These external references are used to maintain the accuracy of the generated documentation.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a data processing pipeline that ingests content from external sources (Google's release index and the Flutter git tree) to generate its own instructional references. This behavior introduces a surface for indirect prompt injection.
  • Ingestion points: External data is ingested by scripts/build_flutter_sdk_changelog.py from Google Cloud Storage and local git checkouts of the Flutter framework.
  • Boundary markers: The reference files generated by the skill do not include delimiters or warnings to prevent the agent from following instructions that might be embedded in the processed data.
  • Capability inventory: The skill includes scripts capable of executing shell commands, making network requests, and performing file system operations.
  • Sanitization: While the scripts use structural parsing (JSON) and regular expressions to extract data, they do not specifically sanitize for natural language instructions that could influence the agent when it processes the resulting markdown files.
Recommendations
  • HIGH: Downloads and executes remote code from: https://storage.googleapis.com/flutter_infra_release/releases/releases_macos.json - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 12, 2026, 01:34 AM
Security Audit — agent-trust-hub — flutter-sdk-changelog