bmad-os-audit-file-refs
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a shell command (
find) to generate a list of files for auditing. The command searches withinsrc/bmm,src/core, andsrc/utility, excluding specific paths, which is consistent with the skill's stated purpose. - [INDIRECT_PROMPT_INJECTION]: The skill processes content from untrusted Markdown and YAML files in the repository using subagents, creating a surface for potential injection.
- Ingestion points: Source files identified by the
findcommand are passed to subagents which use theReadtool to ingest the file content. - Boundary markers: The instructions tell subagents to skip fenced code blocks, but the skill lacks robust delimiters or explicit instructions to ignore adversarial text that might be embedded within the file content.
- Capability inventory: The skill can execute shell commands, spawn subagents, read files, and generate reports based on the aggregated data.
- Sanitization: There is no evidence of content sanitization or validation to prevent adversarial instructions within the source files from influencing the subagents' logic.
Audit Metadata