custom-plugin-flutter-skill-database

Pass

Audited by Gen Agent Trust Hub on Mar 19, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The SyncEngine class in SKILL.md demonstrates a synchronization pattern that fetches data from remote sources using remote.getUpdates(). This architecture introduces a surface for indirect prompt injection where malicious instructions embedded in remote data could influence the agent's behavior during the merge process. Ingestion points: remote.getUpdates() in SKILL.md. Boundary markers: Absent. Capability inventory: Local database persistence and storage management via this.local.upsert. Sanitization: No sanitization or validation of the remote data structure is shown in the examples.
  • [SAFE]: The skill includes legitimate implementations for Hive, SQLite (sqflite), and SharedPreferences. It also demonstrates secure encryption practices using the encrypt package with secure random keys. All external service references, such as Firebase Firestore and Supabase, target well-known technology providers. While the metadata description claims '1800+ lines' of code which is inconsistent with the actual content, this appears to be a descriptive discrepancy rather than a malicious obfuscation attempt.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 19, 2026, 08:50 AM