ask-matt
免费Ask which skill or flow fits your situation. A router over the skills in this repo.
中文简介正在准备中。当前显示原文说明。
Execute approved, bounded work in an isolated implementation conversation and return a structured evidence receipt. Use with inherited history or an explicitly identified source; no particular planning, testing, or review skill is required.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Implement one approved unit of work in the isolated execution conversation. Treat its approved source as authoritative, keep implementation detail here, and return a compact receipt that can be pasted back.
Do not reopen settled product questions. Pause only when the approved source is internally contradictory, conflicts materially with the repository, lacks a required external contract, or requires new authority.
Inspect repository instructions, branch, HEAD, worktree status, recent relevant history, and the current implementation at the agreed seam. Record the actual pre-implementation HEAD as the review fixed point.
Send a short execution lock before substantive edits:
Executing: <one-sentence outcome>
Source: <approved conversation, issue, file, or spec>
In scope: <compact list>
Out of scope: <compact list>
Validation: <primary seam and required gates>
External authority: <what is and is not authorized>
Create or switch to an appropriate local branch when needed. Never discard, overwrite, or absorb unrelated dirty or untracked work. If the requested baseline has drifted, determine whether the spec is still applicable and record the current fixed point; stop if drift changes product behavior or acceptance criteria.
Invoking this skill authorizes in-scope local implementation and validation. It does not by itself authorize committing, pushing, opening or merging a review, changing tracker state, deploying, writing production data, calling costly real services, or messaging people. Perform those actions only when the approved source or a later user message explicitly authorizes them. Never broaden one authorization into another.
When blocked, exhaust safe read-only investigation and local validation first. Then stop with evidence and the smallest decision or authority needed from the planning thread.
If the work proves too large for one reliable implementation context, do not continue by silently dropping history. Return a partial receipt, preserve the worktree, and recommend splitting or compiling the remaining slice into a portable goal.
Do not write a handoff file unless requested. End with one copy-pasteable block:
SPEC EXECUTION RECEIPT
- Conclusion: completed / partially completed / blocked
- Spec source:
- Review fixed point:
- Acceptance criteria: <each criterion with pass/fail and evidence>
- Main changes:
- Changed files:
- Branch / commit / review:
- Validation results:
- Review findings:
- Not validated or not executed:
- Risks and remaining work:
- Planning-thread decision needed:
- Final worktree state:
- External effects: <push, deploy, tracker, data, real services, messages>
- Goal / spec quality: <optional: accurate / criteria-too-vague / criteria-wrong / missing-constraint / over-scoped — plus one sentence>
Keep the receipt concise but evidence-bearing. Include exact commands, counts, identifiers, and links when they materially prove completion. Never claim a real environment, deployment, or external action that was not verified. Redact credentials, tokens, cookies, personal data, and sensitive environment identifiers before the receipt leaves the execution thread.
Goal / spec quality is optional and is not a completion condition. The execution agent may leave it blank. The planning thread or the user fills it after comparing the receipt with the actual diff.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Ask which skill or flow fits your situation. A router over the skills in this repo.
中文简介正在准备中。当前显示原文说明。
Hand the current conversation off to a fresh background agent that picks up the work immediately.
中文简介正在准备中。当前显示原文说明。
Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes: Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/spec asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".
中文简介正在准备中。当前显示原文说明。
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
中文简介正在准备中。当前显示原文说明。
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.
中文简介正在准备中。当前显示原文说明。
Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a GLOSSARY.md, or recording or editing an ADR.
中文简介正在准备中。当前显示原文说明。