bmad-customize
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata and configuration data from other installed skills, which could potentially contain malicious instructions.
- Ingestion points: The
scripts/list_customizable_skills.pyscript readsSKILL.mdfrontmatter descriptions andcustomize.tomlblocks from all subdirectories within the identified skills roots. - Boundary markers: The instructions do not prescribe the use of safety delimiters or specific 'ignore' directives when the agent displays or processes the aggregated skill descriptions.
- Capability inventory: The skill possesses capabilities to write TOML and Markdown files to the local project structure and execute scripts via
uv run. - Sanitization: The discovery logic performs minimal string sanitization (such as quote stripping) on the ingested external descriptions.
- [COMMAND_EXECUTION]: The skill utilizes
uv runto execute local Python scripts designed for skill discovery and configuration resolution. These scripts are provided as part of the skill package or are part of the target project's BMad infrastructure. - [EXTERNAL_DOWNLOADS]: The instructions suggest that users install the BMad framework using
npxfrom the vendor's official GitHub repository (bmad-code-org/BMAD-METHOD) if the environment is not already set up. This uses standard package managers and points to the authoritative vendor source.
Audit Metadata