blacksmith-testbox

Pass

Audited by Gen Agent Trust Hub on May 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches the Blacksmith CLI installer from the official get.blacksmith.sh domain.
  • [REMOTE_CODE_EXECUTION]: The installer script is executed by piping a remote download directly to the shell (curl | sh). This is the documented and primary method for setting up the Blacksmith tool required for the skill's functionality.
  • [DATA_EXFILTRATION]: The skill's core function involves synchronizing local Git-tracked files to remote virtual machines managed by Blacksmith.sh. This process is designed to follow standard CI/CD practices by excluding gitignored files (like .env), minimizing the risk of accidental secret exposure.
  • [COMMAND_EXECUTION]: The skill executes shell commands locally using the blacksmith CLI and various project-specific tools (e.g., pnpm, go test, pytest), as well as executing arbitrary commands within the remote environment.
  • [PROMPT_INJECTION]: The skill processes local repository data, creating a surface for indirect prompt injection where malicious instructions could be embedded in synced files.
  • Ingestion points: Local repository files synced to the remote environment (SKILL.md).
  • Boundary markers: Absent; there are no instructions to the agent to ignore instructions embedded within the synced files.
  • Capability inventory: Remote command execution and file synchronization using the blacksmith CLI (SKILL.md).
  • Sanitization: Absent; the skill does not implement validation or sanitization of file contents before synchronization.
Audit Metadata
Risk Level
SAFE
Analyzed
May 2, 2026, 01:00 PM
Security Audit — agent-trust-hub — blacksmith-testbox