AidenAI has introduced a bundled, project-specific self-knowledge reference in PR #624 to help agents answer questions about their own environment accurately. This change addresses the issue where agents produce confident but ungrounded answers by filling gaps with plausible guesses rather than verifiable facts.
- The skill enforces a hard distinction between acknowledgement (a loggable event) and verified completion, requiring specific evidence for task success.
- It covers hardware ownership, runtime routing, device-bridge behavior, board configuration, setup, observation, verification, and recovery boundaries.
- Scoping is applied per-area rather than using a generic system description to ensure accuracy across different subsystems.
- Validation was structural, including lint checks, YAML frontmatter validation, and focused skill-loader tests, confirming the content is formatted and loadable.
This approach helps agent systems operating across multiple subsystems provide honest answers that depend on which specific subsystem is being queried.