The Agent Skills Directory

[SAFE]: A thorough analysis did not identify any malicious patterns, prompt injection attempts, or unauthorized access to sensitive credentials.
[EXTERNAL_DOWNLOADS]: The skill references documentation and educational resources from the official bmad-code-org GitHub repository and website. It also points to well-known industry documentation for Playwright, Jest, Cypress, and Vitest. These connections are functional for the learning objective and target reputable sources.
[COMMAND_EXECUTION]: The skill executes a local customization resolver script (resolve_customization.py) and internal workflows (advanced-elicitation, party-mode) provided by the platform. These operations are restricted to the local environment and are standard for the agent's initialization and interactive features.
[DATA_EXFILTRATION]: Learner data, including progress tracking and session notes, is stored within local project directories. There is no evidence of exfiltration of sensitive files (~/.ssh, ~/.aws, etc.) or network transmission of personal data.
[SAFE]: The attack surface for indirect prompt injection was assessed as standard for this type of application.
Ingestion points: The skill loads project facts from {project-root}/**/project-context.md and user progress from YAML files.
Boundary markers: Explicit delimiters for ingested facts are not defined in the loading instructions.
Capability inventory: The agent can perform local file reads/writes for state management and execute pre-defined internal scripts/workflows.
Sanitization: Content from context files is loaded verbatim into the session context without specific filtering.

bmad-teach-me-testing