xonecas.io
64875
Mc

mini-coder / commit e773c98

docs: review and improve system prompt, AGENTS.md, and skill descriptions

- Fix typos: propper→proper, Inpirations→Inspirations, propertary→proprietary,
  cronological→chronological, hierchy→hierarchy, accross→across, allways→always,
  ANSII→ANSI, limitted→limited, Tuncates→truncates, targetted→targeted
- Tighten AGENTS.md wording for token efficiency (~4 lines saved)
- Condense context handling and tool surface sections
- Add 'Triggers on:' keywords to all skill descriptions per agentskills.io
  best practices for better trigger matching
- Add 'Use when' clause to debug and mini-coder skill descriptions
- Simplify system prompt mc-edit guideline wording
- Update test assertion to match new wording

xonecas2026-03-21 e773c987d3b02c8df60a35e8dea2edacbe421461 parent cf78df3

7 files changed

.agents/skills/audit/SKILL.md+1−1
@@ -1,6 +1,6 @@
11 ---
22 name: audit
3-description: "Full codebase audit — code review, multi-provider UI testing, and combined report. Use when the user asks to audit, test across providers, or validate the full implementation."
3+description: "Full codebase audit — code review, multi-provider UI testing, and combined report. Use when the user asks to audit, test across providers, or validate the full implementation. Triggers on: audit, full review, test providers, validate implementation."
44 compatibility: "Requires tmux, bun, git, and active provider credentials (Anthropic OAuth and/or Opencode zen)"
55 ---
66
.agents/skills/debug/SKILL.md+1−1
@@ -1,6 +1,6 @@
11 ---
22 name: debug
3-description: Debug mini-coder issues by analyzing session logs and database entries. Access error logs, API call logs, and session history from the database.
3+description: "Debug mini-coder issues by analyzing session logs and database entries. Use when investigating errors, API failures, token usage, or session history. Triggers on: debug, logs, errors, API logs, session history, token usage."
44 allowed-tools: shell
55 ---
66
.agents/skills/mini-coder/SKILL.md+1−1
@@ -1,6 +1,6 @@
11 ---
22 name: mini-coder
3-description: How to use the `mc` CLI coding agent — interactive sessions, one-shot prompts, slash commands, file editing with mc-edit, and orchestrating subagents for complex tasks.
3+description: "How to use the `mc` CLI coding agent — interactive sessions, one-shot prompts, slash commands, file editing with mc-edit, and orchestrating subagents for complex tasks. Use when helping users operate mc, or when acting as an mc agent yourself."
44 ---
55
66 # mini-coder
.agents/skills/release/SKILL.md+1−1
@@ -1,6 +1,6 @@
11 ---
22 name: release
3-description: "Create a new tagged npm release (patch/minor/major). Use when the user asks to release, publish, bump version, or cut a new version."
3+description: "Create a new tagged npm release (patch/minor/major). Use when the user asks to release, publish, bump version, or cut a new version. Triggers on: release, publish, bump, version, tag."
44 compatibility: "Requires git, bun, and npm with publish credentials"
55 ---
66
AGENTS.md+27−31
@@ -13,7 +13,7 @@
1313 - Do not inline `import` calls. Don't duplicate code. Don't leave dead code behind.
1414 - Don't re-implement helpers/functionality that already exists, always consolidate.
1515 - Don't add complexity for backwards compatibility, it's preferable to break compatibility and keep the code simple.
16-- Don't make random test files, if you need to test something, write a propper unit test.
16+- Don't make random test files, if you need to test something, write a proper unit test.
1717 - If you make temp files, clean them up when you are done.
1818 - Use Conventional Commits formatting for commit messages. When you commit, include the whole diff unless told otherwise.
1919 - Always use your superpowers skills effectively.
@@ -22,24 +22,23 @@
2222
2323 mini-coder - `mc`
2424
25-A coding agent via a augmented shell prompt, small and fast that doesn't get in the way
25+An augmented shell prompt coding agent — small, fast, stays out of the way.
2626
27-### Inpirations:
27+### Inspirations:
2828
29-- Pi agent: https://github.com/badlogic/pi-mono/tree/main/packages/coding-agent amazing tool, like a swiss army knife of coding agents.
30-- Claude Code: https://claude.com/product/claude-code clean UI, propertary software
31-- Codex: https://github.com/openai/codex Powerful and open source, focused on heavy open-ai integration
32-- Opencode: https://opencode.ai/ The open source standard. Similar to oh-my-pi but with a different aesthetic
29+- Pi agent: https://github.com/badlogic/pi-mono/tree/main/packages/coding-agent
30+- Claude Code: https://claude.com/product/claude-code
31+- Codex: https://github.com/openai/codex
32+- Opencode: https://opencode.ai/
3333
3434 ### Philosophy:
3535
36-**Fast and performant**, instant feedback shell like output. Performance first style second.
36+**Performance first, style second.** Instant feedback, shell-like output.
3737
38-Very focused on dev flow, simple setup steps and fast to coding.
39-Elegant output that focus on the conversation with the agent and their actions. Ensure an accurate cronological conversation log that is very readable for the user. Keep a visible hierchy of actions -> results and assistant reasoning and responses.
40-Accurate and reliable and fast core set of features.
41-Community oriented, all of the inspirations have their own dotfile format, try to follow the existing
42-conventions and not introduce more specs.
38+Focused on dev flow — simple setup, fast to coding.
39+Elegant output focused on the conversation and agent actions. Accurate chronological log with clear hierarchy: actions → results → reasoning → responses.
40+Accurate, reliable, fast core features.
41+Community oriented — follow existing dotfile conventions, don't introduce new specs.
4342
4443 ### Features:
4544
@@ -51,7 +50,7 @@ conventions and not introduce more specs.
5150
5251 - Auto discovery of providers via ENV (Example: OPENCODE_API_KEY), or local servers (Example: ollama)
5352 - `/login` shows OAuth login status. `/login <provider>` starts browser-based OAuth login. `/logout <provider>` clears saved tokens. Currently supports Anthropic. (Both pi and opencode support this too)
54- - `/model` (alias `/models`) command allows the user to pick a model from connected providers. As well as thinking effort for the model if supported. Selection persists accross sessions. Must have autocomplete support for ease of use when changing models
53+ - `/model` (alias `/models`) picks a model from connected providers, including thinking effort if supported. Selection persists across sessions with autocomplete.
5554
5655 - Session management and command to create new/resume/list with local sqlite file.
5756 - `/session` command to interactively manage sessions as well as cli flags.
@@ -66,24 +65,21 @@ conventions and not introduce more specs.
6665 - Shell like autocomplete for filepaths, `@` embeds a file into the prompt and autocompletes filepaths
6766
6867 - Context handling:
69- - Elegantly handles hitting max context size for the model using regular per step pruning. Stops with a conversation summary if max context is reached.
70- - No max steps or max tool calls. user can interrupt, this is not needed.
71- - Support prompt caching via the used sdks.
72- - Rolling context pruning with a balanced default. (Must not break caching!)
73- - Clear and accurate token tracking.
74- - Recover from errors returning the user to the prompt with clear messaging
75- - `/undo` removes the last turn from conversation history, it does not restore filesystem state.
76-
77-- Shell-first tool surface for LLMs: `shell`, `listSkills`, and `readSkill`, plus connected MCP tools and optional web tools when configured. Keep the core surface small and easy to use so it actually reduces friction for the LLMs.
78- - File work should follow a simple flow: inspect with shell, mutate with `mc-edit`, verify with shell.
79- - Large shell outputs should be truncated automatically for the llm, to avoid context explosions and encourage targetted reads.
80- - `mc-edit` should stay narrow and reliable: exact-text edits only, deterministic failures on stale or ambiguous state, diff and machine-friendly output while staying human readable.
81- - Optional `webSearch` and `webContent` tools when `EXA_API_KEY` is set.
82- - Cross model system prompt with clear descriptions of the expected dev flow and guidance as a code agent focused on shell first.
68+ - Rolling context pruning per step; stops with a conversation summary at max context. Must not break prompt caching.
69+ - No max steps or tool call limits — user can interrupt.
70+ - Clear, accurate token tracking.
71+ - Recover from errors gracefully, returning user to the prompt.
72+ - `/undo` removes the last turn from history (does not restore filesystem).
73+
74+- Shell-first tool surface: `shell`, `listSkills`, `readSkill`, plus connected MCP tools and optional web tools (`EXA_API_KEY`). Keep it small.
75+ - Inspect with shell → mutate with `mc-edit` → verify with shell.
76+ - Auto-truncate large shell outputs for the LLM to avoid context explosions.
77+ - `mc-edit`: exact-text edits only, deterministic failures on stale/ambiguous state.
78+ - Cross-model system prompt with clear dev flow guidance.
8379
8480 - Other Commands in CLI prompt:
8581 - `/reasoning` toggles display of model reasoning output.
86- - `/verbose` toggles output truncation. Tuncates large outputs keeping start and end sections and truncating the rest when verbose is off. This controls truncation for the user output, it's a visibility UI concern
82+ - `/verbose` toggles output truncation. When off, large outputs keep start/end sections and truncate the middle. UI-only visibility concern.
8783 - `/review` reviews recent changes via a global custom skill installed at app start (`~/.agents/skills/review.md`), and can be customized or shadowed locally.
8884
8985 - Connect to MCP servers over Streamable HTTP / SSE fallback or stdio.
@@ -91,7 +87,7 @@ conventions and not introduce more specs.
9187
9288 ### UI/Output
9389
94-- Use the 16 ANSII colors so the coloring is allways inherited from the terminal theme (https://jvns.ca/blog/2024/10/01/terminal-colours/)
90+- Use the 16 ANSI colors so coloring is always inherited from the terminal theme (https://jvns.ca/blog/2024/10/01/terminal-colours/)
9591 - Use pi and claude code as visual inspirations.
9692 - Banner at app start, show found configs and context files
9793 - history log (like a terminal output scrolls away, pushing the input prompt down)
@@ -105,7 +101,7 @@ conventions and not introduce more specs.
105101
106102 - All things Bun.js, runtime, package manager. https://bun.com/docs
107103 - Multiprovider support via https://ai-sdk.dev/docs/introduction
108-- Colored output with https://github.com/sindresorhus/yoctocolors limitted but fast!
104+- Colored output with https://github.com/sindresorhus/yoctocolors — limited but fast.
109105
110106 ### Repo structure
111107
src/agent/system-prompt.test.ts+1−1
@@ -101,7 +101,7 @@ describe("buildSystemPrompt", () => {
101101 expect(prompt).toContain(
102102 "Inspect code and files primarily through shell commands",
103103 );
104- expect(prompt).toContain("use shell commands that invoke `mc-edit`");
104+ expect(prompt).toContain("invoke `mc-edit` via shell");
105105 expect(prompt).toContain("check the available skills list below");
106106 expect(prompt).toContain("never guess unknown state");
107107 });
src/agent/system-prompt.ts+1−1
@@ -54,7 +54,7 @@ Current date/time: ${sessionTimeAnchor}
5454 Guidelines:
5555 - You are a capable senior engineer. Proactively gather context and implement — work the problem, not just the symptom. Prefer root-cause fixes over patches.
5656 - Inspect code and files primarily through shell commands. Use temp files for large content to avoid filling your context window.
57-- For file edits, use shell commands that invoke \`mc-edit\`. Prefer small, targeted edits over full rewrites so diffs stay reviewable.
57+- For file edits, invoke \`mc-edit\` via shell. Prefer small, targeted edits over full rewrites so diffs stay reviewable.
5858 - Make parallel tool calls when the lookups are independent — this speeds up multi-file investigation.
5959 - Before starting work, check the available skills list below. If any skill's description/triggers match your current task, load it with \`readSkill\` and follow its instructions before proceeding.
6060 - Keep it simple: DRY, KISS, YAGNI. Avoid unnecessary complexity.