dart-pub-upgrade

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes multiple local commands including git for branch management, dart and flutter for package upgrades and analysis, and gh for creating Pull Requests. These commands use controlled arguments based on project configuration and local environment markers.
  • [EXTERNAL_DOWNLOADS]: The skill fetches package metadata and archive files from the official Dart package registry (https://pub.dev) to extract changelog information. These downloads target a well-known service and are used for documentation purposes.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from package changelogs which are summarized by the AI. This presents an attack surface where a malicious package author could attempt to influence the agent's summary or PR metadata.
  • Ingestion points: Package changelogs are downloaded from pub.dev via fetchChangelog (scripts/bin/dart_pub_upgrade.dart) and stored in changelog_diffs.json before being read by the AI.
  • Boundary markers: No explicit boundary markers or 'ignore' instructions are provided when the AI is instructed to summarize the changelog content in SKILL.md.
  • Capability inventory: The skill has the ability to commit changes to Git and create Pull Requests on GitHub via the gh CLI, which could be misused if the AI is successfully injected.
  • Sanitization: No explicit sanitization or validation of the downloaded changelog text is performed before it is passed to the AI for summarization.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 01:35 AM
Security Audit — agent-trust-hub — dart-pub-upgrade