skill-up
Aid agent skill creation or modification. Use when creating a new skill or changing an existing one. Use whenever writing any agent instructions, whether to a file or directly in a reply.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- skill-up
- description
- Aid agent skill creation or modification. Use when creating a new skill or changing an existing one. Use whenever writing any agent instructions, whether to a file or directly in a reply.
03 SKILL.md 正文
<straw-dog question="q-mzn7ek">Mechanism: not yet</straw-dog>
Your goal is to aid agent skill creation or modification.
Writing rules — to the mechanism shape, which produces instructions, a skill is a record of the kind what must always hold, and these are that kind's invariants for one:
- A skill is instruction, not story. Be precise and concise. Prefer umbrella terms to enumeration, unless can be interpreted wrong in context of the skill.
- State everything in definitive form — no evolution logic, no decision explanations.
- Write a filter, a fixed form or a must only where it always holds; where judgement decides, leave it to the model rather than overfit the instruction.
Before writing, browse the existing skill set and match its structure and vibe — frontmatter shape, file layout, linking style, tone, altitude. Derive the set's conventions from the set itself; do not impose foreign ones.
Verify the new or changed skill against the set:
- It does not overlap an existing skill, unless the overlap is intentional — then name it and link the owning skill.
- It does not contradict any existing skill.
- It does not misfit the set — wrong altitude, wrong output location, conventions the set does not use.
A fact the set already states in one skill is referenced from there, not restated.
Weigh the whole skill on every change: say what the change removes before what it adds, and compare the body's size before and after. A body that has only grown is read whole for what restates another home, what a script already does for the reader, and what is a record's format, which goes to a format file beside it.
The frontmatter description narrates the use case — the skill's moments —, not the implementation, and stays in line with the skill's intent; the body holds what to do at them and never restates when. Capturing skill intent and usage is critical — /align when in tiniest doubt.
Project-local divergence
A skill the harness ships carries a project fact only in the local block its installer writes; its body and appendices carry none. A project's own skill cites the project's documentation for what it holds.
Merging a reference corpus
When asked to merge the skill corpus, ask which corpus to merge from. Port body changes, fixing defects as you port; leave every appendix file and installed block as it stands. A skill present in only one corpus is either local by intent or not yet ported — ask which.
Installed from other mechanisms
<installed by="mechanism-shape"> R5 Before writing a rule into AGENTS.md, find the mechanism whose work it governs. Land a mechanism's rule in AGENTS.md only as an installed block from its rules file. While the mechanism is undeclared, write the rule by hand and wrap it as a straw dog bound to the question of what the mechanism owns, as you write it.
R6 A skill is an instruction file of exactly one mechanism and opens its body with
Mechanism: <slug>. A skill no mechanism names opens it with Mechanism: then not yet,
wrapped as a straw dog bound to the question of what the skill owns, or unowned by design with
its reason. A script, test or hook a mechanism owns opens with # Mechanism: <slug>, after a
shebang where there is one. Write the line as you add the file; no list holds it.
R8 Name no skill of a mechanism whose state is installed in another skill's own text: what it asks
of that skill reaches it as an installed block from its rules file.
</installed>
<installed by="local">
L10 What the harness ships names under docs/ only where a mechanism keeps a kind of record — the
directory, or the one file that is a record — never a record inside one; a question only as a
straw dog's binding or a TODO q-N; a ticket never.
</installed>
04 同仓库更多 Skills
verify
Use after an implementation pass, or to verify a landed slice before close. Whole verification of landed work: ticket, RFC, governing docs, and the project's verification set.
ticket
Use to decompose a parent work into independently-workable child tickets. Use to mint ticket(s), possibly a single one, when planning implementation without one.
tdd
Test-driven development with red-green-refactor loop. Use when user wants to build features or fix bugs using TDD, mentions "red-green-refactor", wants integration tests, or asks for test-first development.
setup-devops
Set up the project's engineering scaffolding, from toolchain to CICD and observability. Use when initializing a repo or filling a setup gap.
review-architecture
Explore the codebase for module-deepening opportunities — architectural improvements that raise testability and navigability. Use when asked to improve or refactor the architecture.
recall
Determine where the project stands and what to work on next, with the decisions already taken surfaced alongside the questions genuinely still open. Use when asked what is next, where things stand, or to continue or pick up work.