skills/keboola/ai-kit/test-component/Gen Agent Trust Hub

test-component

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and process external data files from the Keboola platform (stage_output.zip), creating an attack surface where instructions embedded in these files could influence agent behavior.
  • Ingestion points: The agent is directed to unzip stage_output.zip, read and parse config.json, and process manifest files in references/vcr-debug-from-platform.md.
  • Boundary markers: Absent. There are no explicit delimiters or instructions for the agent to ignore potentially malicious commands embedded in the processed configuration or data files.
  • Capability inventory: The skill has access to Bash, Read, Write, and Edit tools, enabling it to write files, modify source code, and execute arbitrary Python code via uv run or pytest.
  • Sanitization: Partial. The skill provides regex-based scripts to redact [hidden] placeholders and uses VCR sanitizers to strip secrets from network recordings, but it does not sanitize instructions in the input data.
  • [COMMAND_EXECUTION]: The skill relies on the Bash tool to perform developer tasks, including environment setup, file management, and running tests.
  • Evidence: Instructions in SKILL.md and references/vcr-quickstart.md use commands like uv sync, uv run python -m keboola.datadirtest scaffold, and pytest.
  • Evidence: references/vcr-debug-from-platform.md uses shell commands like unzip, mkdir, and mv to manipulate platform debug files.
  • [EXTERNAL_DOWNLOADS]: The skill downloads and installs dependencies from official and vendor-related repositories.
  • Evidence: Fetches keboola.datadirtest and keboola.component from GitHub and standard package registries during project synchronization.
  • Evidence: Standard development tools like pytest and freezegun are listed as requirements in pyproject.toml.
  • [DYNAMIC_EXECUTION]: The skill generates and executes Python scripts at runtime to automate the fixing of configuration files.
  • Evidence: In references/vcr-debug-from-platform.md, a bash heredoc is used to create and run a Python script (`python3
  • <<'EOF' ... EOF) that uses regex and JSON modules to modify config.json`.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 07:50 AM