NYC

db-wal-recovery

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • COMMAND_EXECUTION (LOW): The skill suggests executing shell commands like xxd, hexdump, and ls to inspect database files. These commands are necessary for the skill's primary purpose of forensic data recovery but involve subprocess execution on potentially untrusted file paths.
  • INDIRECT PROMPT INJECTION (LOW): The skill is designed to ingest and analyze untrusted data from SQLite WAL files without explicit boundary markers. 1. Ingestion points: SQLite database and WAL files (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: ls, stat, file, readlink, xxd, hexdump (SKILL.md). 4. Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:05 PM