init
Register the developer-plenger skill pack by writing AGENTS.md at the repo root — the pack's workflow, artifacts, skill rules and task-state contract — and create the empty context.md that every later skill updates. Use for /init, when onboarding a repository to the pack, or when AGENTS.md is missing or has drifted from the pack contract. Asks the user nothing and inspects no project facts; it does not create the project, pick a stack, or describe what the project is.
- License 许可
- MIT
- 最近更新
- 2026-10-29
- Stars
- ★ 3
- Forks
- 0
01 前言 Front Matter
- name
- init
- description
- Register the developer-plenger skill pack by writing AGENTS.md at the repo root — the pack's workflow, artifacts, skill rules and task-state contract — and create the empty context.md that every later skill updates. Use for /init, when onboarding a repository to the pack, or when AGENTS.md is missing or has drifted from the pack contract. Asks the user nothing and inspects no project facts; it does not create the project, pick a stack, or describe what the project is.
02 资源文件 2
- 参考文档agents/openai.yaml372 B
- 参考文档references/agents-template.md12.6 KB
03 SKILL.md 正文
Init — register the skill pack
/init writes two files at the repo root so every later skill knows which skills exist, what each one produces, where each artifact lives, and what the project currently is:
AGENTS.md— the pack contract. Identical in every project; it is the pack's rules, not a description of the project.context.md— the project's memory, created empty: thirteen sections and nothing in them, because nothing has been planned yet and/initinspects nothing.
The run is: read what exists, compare it against the contract, write if it does not conform, report. /init asks the user nothing — there is no project-specific content to gather. It never inspects the codebase to infer a stack, a purpose or an architecture; those are project facts and belong to /plan, which records them in context.md and in specs/NN-<plan-slug>/spec.md.
Read references/agents-template.md before step 2 for the AGENTS.md contract — the canonical text, the conformance checklist, the merge rule for an existing file, and the CLAUDE.md dispositions. Read skills/plan/references/context.md before step 5 for the context.md shape that step writes.
Steps
1. Read what already exists
Read, in this order, and only if present:
AGENTS.mdat the repo rootCLAUDE.mdat the repo rootcontext.mdat the repo root
Done when: you know whether AGENTS.md exists, whether context.md exists, and whether a CLAUDE.md needs a disposition decision.
2. Compare AGENTS.md against the contract
Open references/agents-template.md §3 and check the existing file, section by
section, against the canonical text in §2. There is nothing to detect and nothing
to infer: the expected content is fixed, so this is a comparison.
If AGENTS.md is absent, this step is trivially "non-conformant" and you go
straight to writing it.
Done when: you can name every section that is missing or drifted, or state that none is.
3. Write AGENTS.md
- Absent — write the canonical text from
references/agents-template.md§2 verbatim. - Present and conformant — write nothing.
- Present and drifted or incomplete — write the canonical version into place, following the merge rule in §4: never clobber content that is not a pack section, never reshuffle the user's own sections.
Never resolve a section by guessing project facts. Nothing in this file is project-specific, so there is nothing to infer.
Done when: AGENTS.md at the repo root matches the contract, or was already matching.
4. Resolve CLAUDE.md, if present
If CLAUDE.md exists, apply the dispositions in references/agents-template.md
§4 and say plainly which file is canonical and whether the other becomes a
symlink or a short pointer. If it does not exist, do nothing here.
Done when: the user has been told which files you wrote, and — when a CLAUDE.md exists — which file is canonical and what happens to the other.
5. Create context.md, or reset its cursor
Follow skills/plan/references/context.md for the exact shape.
- Absent — write the thirteen sections, all empty, with
Current Phaseset toINIT. Empty means blank, notunknown:/planis the run that fills this file, and a placeholder here invites a later skill to treat a guess as a fact. - Present — set
Current PhasetoINITand change nothing else. The project's facts belong to/plan.
Create no directory here. specs/ is created by /plan, and a fresh clone should have no specs/ tree until a plan is actually written.
Done when: context.md exists with the thirteen sections and Current Phase: INIT, or was already present with only its cursor changed.
6. Report
Tell the user: whether AGENTS.md was created, repaired, or already conformant; whether context.md was created or left in place; any non-pack content you preserved and where it belongs if it is still sitting in AGENTS.md; any decision taken about CLAUDE.md; and the next command to run — /plan <idea>.
Done when: the user knows the state of both files and what to run next.
04 同仓库更多 Skills
writing-shape
Writing, exploit: shape raw material into an article, paragraph by paragraph.
writing-fragments
Writing, explore: mine raw fragments, no structure yet.
writing-for-agents
Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.
writing-beats
Writing, exploit; assemble raw material into a journey of beats, grounding each term before a beat leans on it.
wizard
Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.
wayfinder
Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.