agent-file-synchronization

Pass

Audited by Gen Agent Trust Hub on Aug 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute a local utility script, scripts/sync_instruction_files.py, which contains the logic for mirroring instruction files while preserving platform-specific sections.\n- [PROMPT_INJECTION]: The skill propagates content from CLAUDE.md to other agent instruction files, which presents a surface for indirect prompt injection if an untrusted user is able to modify the source instruction file.\n
  • Ingestion points: Content from the CLAUDE.md file in the repository root.\n
  • Boundary markers: Uses a static anchor string ('Behavioral guidelines to reduce common LLM coding mistakes...') to identify the shared instruction body.\n
  • Capability inventory: Modified files include .github/copilot-instructions.md, GEMINI.md, and AGENTS.md via the Write tool.\n
  • Sanitization: Content is replicated directly without automated sanitization or filtering of the instructions being mirrored.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 9, 2026, 12:57 PM
Security Audit — agent-trust-hub — agent-file-synchronization