claude-sync

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes several shell commands including git, rsync, cp, test, and rm to validate repository state and perform file synchronization operations.
  • [DATA_EXFILTRATION]: The skill accesses and copies files such as .claude/settings.json and .claude/mcp.json, which often store sensitive credentials, API tokens, and internal environment details. While the operations are local, this represents a potential exposure of sensitive secrets.
  • [PROMPT_INJECTION]: The skill presents a risk of indirect prompt injection by copying natural language instruction files (CLAUDE.md, skills) from an external source to the target codebase.
  • Ingestion points: Files read from the user-specified source_path as described in SKILL.md.
  • Boundary markers: No delimiters or protective instructions are used to prevent the agent from obeying instructions embedded in the synced files.
  • Capability inventory: The skill has the ability to perform file writes (shutil.copy2) and execute system commands (git, rsync) within the target repository.
  • Sanitization: No content filtering, escaping, or validation is performed on the data being synced.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 11:08 AM
Security Audit — agent-trust-hub — claude-sync