agenttune

AgentTune is an open (MIT) library of 43 personality tuning files that align an AI agent's interaction style with how a specific user thinks — five systems: MBTI (16), Enneagram (9), DISC (4), Attachment (4), OCEAN/Big Five (10 compositional high/low files). Typical flow: 1. User knows their type → get_tuning(system, slug). Apply the returned Markdown as system-prompt content for the session/project (CLAUDE.md, AGENTS.md, custom instructions, or the API system parameter — the file's YAML front-matter lists exact per-surface install paths under install.surfaces and a verification probe under verify.probe). 2. User doesn't know their type → get_test_spec(test), administer the items inline (bulk-paste all items and ask for a response array), score per the spec's algorithm, then get_tuning with the resulting slug. 3. Tunings layer: one per system, concatenated. On conflict, precedence is OCEAN (measured) > Attachment / DISC / MBTI / Enneagram (categorical). The user connected thi