vault-snapshot

Warn

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a shell script wiki-snapshot.sh (SKILL.md) and involves high-privilege system operations including the use of sudo and the modification of systemctl (SKILL.md).
  • [COMMAND_EXECUTION]: Persistence is established by installing a systemd timer (wiki-snapshot.timer) configured to run every 30 minutes (SKILL.md).
  • [DATA_EXFILTRATION]: The skill performs automated synchronization of data between S3 and GitHub (SKILL.md). These operations involve well-known cloud services.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from a remote S3 store for propagation to a Git repository, creating a risk of indirect prompt injection. Mandatory Evidence Chain: 1. Ingestion points: S3 direct store content (SKILL.md); 2. Boundary markers: Absent; 3. Capability inventory: rclone sync, git push, and shell execution (SKILL.md); 4. Sanitization: Stripping of delete-intent tombstone paths (SKILL.md).
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 27, 2026, 11:30 AM
Security Audit — agent-trust-hub — vault-snapshot