skills/bmad-labs/skills/bmad-agent-pm/Gen Agent Trust Hub

bmad-agent-pm

Warn

Audited by Gen Agent Trust Hub on May 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a Python script located at {project-root}/_bmad/scripts/resolve_customization.py during its activation process. This involves executing code from a computed path within the project environment.
  • [COMMAND_EXECUTION]: The activation workflow includes steps to execute arbitrary entries defined in {agent.activation_steps_prepend} and {agent.activation_steps_append}. While empty by default, this mechanism allows for the execution of arbitrary shell commands or instructions if configuration overrides are present in the project structure.
  • [DATA_EXFILTRATION]: The skill implements broad file system access using glob patterns (e.g., file:{project-root}/**/project-context.md) to load contents as persistent facts. This ingestion of project-level data without clear boundary markers or sanitization creates an attack surface for indirect prompt injection.
  • Ingestion points: {project-root}/**/project-context.md in SKILL.md (Step 4).
  • Boundary markers: Absent; the content is loaded directly as foundational context.
  • Capability inventory: Local Python script execution (Step 1), arbitrary command execution via activation steps (Steps 2 & 7), and invocation of other skills via the menu system.
  • Sanitization: None detected for the ingested file content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 10, 2026, 05:25 PM
Security Audit — agent-trust-hub — bmad-agent-pm