skills/binhmuc/autobot-review/repomix/Gen Agent Trust Hub

repomix

Pass

Audited by Gen Agent Trust Hub on Oct 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and package external repositories (local or remote) to be used as context for an AI agent. This creates a surface for indirect prompt injection, where malicious instructions embedded in a target codebase could be interpreted by the agent after the code is packaged.
  • Ingestion points: The scripts/repomix_batch.py script accepts repository paths and URLs via command-line arguments and JSON configuration files.
  • Boundary markers: While the repomix tool uses XML/Markdown tags to structure the output, the skill lacks specific instructions for the agent to treat the packaged content as untrusted data.
  • Capability inventory: The skill can read local files and fetch remote repositories using the repomix CLI via subprocess.run.
  • Sanitization: The skill includes a secret scanning check (Secretlint), but it does not provide sanitization or filtering for potential prompt injection patterns in the processed data.
  • [COMMAND_EXECUTION]: The scripts/repomix_batch.py script dynamically constructs and executes commands to run the repomix CLI.
  • The script uses subprocess.run with a list of arguments. While this avoids shell injection, it directly incorporates user-supplied repository paths and output filenames as command arguments.
  • The script provides an option (--no-security-check) that allows users to disable the tool's integrated secret scanner.
  • [DATA_EXFILTRATION]: The skill contains logic and artifacts that risk exposing sensitive system information.
  • The binary file scripts/.coverage (a SQLite database) contains code coverage metadata that leaks absolute file paths (e.g., /mnt/d/www/claudekit/...), revealing the author's internal directory structure.
  • The EnvLoader class in scripts/repomix_batch.py automatically searches for and loads .env files from several parent directory levels, including the agent's configuration directories (e.g., .claude/.env). This behavior could lead to the unintentional loading and exposure of sensitive credentials to the subprocess environment.
  • [EXTERNAL_DOWNLOADS]: The skill relies on and facilitates the execution of the repomix tool from the npm registry.
  • The repomix_batch.py script uses npx repomix for remote repository processing, which triggers the download and execution of the package from the npm registry at runtime.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 9, 2026, 04:35 AM