skills/rheinmir/setup/sync-template/Gen Agent Trust Hub

sync-template

Warn

Audited by Gen Agent Trust Hub on Jun 24, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches template files and repository branch information from the author's GitHub repository (https://github.com/Rheinmir/setup.git) using curl and the GitHub CLI.
  • [COMMAND_EXECUTION]: Uses shell commands to manage local files, including creating project and global directories (mkdir -p), comparing file contents (diff), and migrating directory structures.
  • [REMOTE_CODE_EXECUTION]: Downloads remote markdown files and installs them into specialized agent directories (~/.claude/skills/, ~/.claude/commands/). Since these files contain instructions the agent interprets and executes, this effectively installs remote code into the agent's global environment.
  • [DATA_EXFILTRATION]: Includes functionality to push local project files to a remote repository ('Upstream sync'). Although the skill contains rules against syncing credentials, it establishes a pathway for transferring local data to an external destination.
  • [PROMPT_INJECTION]: The skill's primary function is to ingest and execute remote instructions, creating an indirect prompt injection surface. Evidence: 1. Ingestion points: Remote GitHub repository via curl. 2. Boundary markers: Absent. 3. Capability inventory: curl, mkdir, printf, cat, gh api, and file writes to global configuration paths. 4. Sanitization: Absent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 24, 2026, 09:06 AM
Security Audit — agent-trust-hub — sync-template