v1.7.2

• Config moved from ~/.mirc to ~/.mi/config.json (overridable via MI_HOME)
• Line count limit now applies to index.mjs only — tools are excluded
• Split dense lines in index.mjs for readability (still 30 LOC)
• Fixed test isolation so user config doesn't leak into the test suite

v1.7.1

Add --help flag support (alongside existing -h).

v1.7.0

~/.mirc config file

Optional JSON config at ~/.mirc — keys become env var defaults. Shell env always wins.

{
  "MODEL": "o3",
  "OPENAI_BASE_URL": "http://localhost:11434",
  "REASONING_EFFORT": "high"
}

Any env var mi reads can be set here: OPENAI_API_KEY, MODEL, OPENAI_BASE_URL, REASONING_EFFORT, SYSTEM_PROMPT.

Zero new lines — still 30 LOC.

v1.6.3

• Add optional REASONING_EFFORT env var support
• Misc chores (readme, splash)

v1.6.2

Full Changelog: v1.6.1...v1.6.2

v1.6.1

Self-Extending Agent

The agent can now write its own tools via the self skill, including tools that spawn sub-agents.

Changes

• Improved tool authoring guidance in self skill with step-by-step procedure
• Added recursive mi example (tools that spawn sub-agents)
• Clarified skill vs tool distinction in new-skill
• Added "Be concise" and agent identity to system prompt
• README: added self-extending and recursive agents features

Verified

• Agent successfully creates new tools following the self skill
• Recursive mi tools (sub-agents) work correctly

v1.6.0

Modular Tools

Tools now live in separate files (tools/*.mjs) and are auto-discovered at startup. This makes it easier to add new tools without modifying the core harness.

Changes

• Extract bash and skill tools into tools/bash.mjs and tools/skill.mjs
• Harness auto-discovers tool modules at startup
• Common Node modules exposed as globals for tool authors
• Add CLAUDE.md symlink to AGENTS.md
• Update self skill with tool authoring guide

Tool Authoring

New tools can be added by creating a .mjs file in the tools/ directory:

export default {
  name: 'mytool',
  description: 'What it does',
  parameters: { type: 'object', properties: { arg: { type: 'string' } }, required: ['arg'] },
  handler: async ({arg}) => 'result'
};

Total line count: 30 (23 harness + 4 bash + 3 skill)

v1.5.1

• feat(skills): bundle new-skill (author SKILL.md procedures) and self (introspect mi's source/config/runtime)
• feat(repl): /reset now acks with a dim ✓ reset line so it's clear the history was cleared
• docs(readme): list all 11 bundled skills (was stale at 6)

v1.5.0

What's new

• Streaming: assistant tokens now stream to stdout via SSE as they arrive, instead of waiting for the full reply. Works in REPL, -p, and stdin modes.
• REPL polish: version banner on startup, error recovery (failed requests pop the user message instead of crashing).

Internals

• run() sends stream: true and iterates res.body, parsing SSE delta events. Content writes to stdout incrementally; tool-call fragments merge by index into one assistant message.
• Test mock server converted to emit SSE.
• 27 → 29 LOC.

v1.4.0

Skills overhaul

All 9 bundled skills improved across 12 iterations:

• Sharper descriptions — every skill now leads with concrete trigger phrases so the model selects the right one
• Cross-references — debug→tdd, tdd→refactor, plan↔tasks wired up at decision points
• Red Flags sections — added to debug, tdd, verify, refactor, review so agents know when to stop and ask
• Concrete examples — review FINDINGS format, debug observation template, verify per-tool output patterns
• File-writing guidance — debug and tdd now warn against echo+\n and sed for multi-line files
• Portability — harness exports MI_PATH; delegate skill uses it instead of hardcoded paths
• plan — explicit /tmp/mi-/plan.md path with copy-pasteable command
• tasks — explicit work loop (pick → mark in-progress → do → mark done); slug collision guidance
• delegate — log-reading strategies (tail, grep, wc); blocked-subagent handling in review

Fixes

• Fixed test suite deleting bundled skills/ directory after test runs
• Harness now sets MI_PATH env var so subagents can spawn further subagents
• .gitignore added; .npmignore tightened