rr-plan
Plan work before any edit. Given a spec (docs/sdlc/specs/<slug>.md), write the programme plan docs/sdlc/plans/<slug>.md — files that change, order of work, risks, proof per requirement, rollback. Given one task, measure its weight with `state.py weight` (six axes, seven lines), take the mode it gives (light, guarded, full), shape the plan by that mode — inline, plan mode, or written — collect every question the plan depends on now, and register the task with `state.py start`, which arms the scope guard. Use on "/rr-plan", "plan this", "plan the implementation", "make the plan from the spec", and whenever /rr-build finds no task in flight.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 3
- Forks
- 0
01 前言 Front Matter
- name
- rr-plan
- description
- Plan work before any edit. Given a spec (docs/sdlc/specs/<slug>.md), write the programme plan docs/sdlc/plans/<slug>.md — files that change, order of work, risks, proof per requirement, rollback. Given one task, measure its weight with `state.py weight` (six axes, seven lines), take the mode it gives (light, guarded, full), shape the plan by that mode — inline, plan mode, or written — collect every question the plan depends on now, and register the task with `state.py start`, which arms the scope guard. Use on "/rr-plan", "plan this", "plan the implementation", "make the plan from the spec", and whenever /rr-build finds no task in flight.
- argument-hint
- <task in one sentence> | <path to docs/sdlc/specs/<slug>.md>
03 SKILL.md 正文
/rr-plan $ARGUMENTS
Plan, never edit: this skill reads code and writes a plan, and the first edit
belongs to /rr-build. Resolve the install root once:
for AI_HOME in "${CLAUDE_CONFIG_DIR:-$HOME/.claude}" "${CODEX_HOME:-$HOME/.codex}"; do
[ -d "$AI_HOME/skills/rr-plan" ] && break
done
STATE="python3 $AI_HOME/skills/ai-task/state.py"
Read docs/sdlc/constitution.md first and cite the principles a decision rests
on as C<n>; if it is absent, say so in one line and go on. $STATE status
first: a task already in flight is finished or abandoned before another starts.
A spec → the programme plan
When $ARGUMENTS is a spec (default: the newest docs/sdlc/specs/*.md):
docs/sdlc/constitution.mdexists, else/rr-setup. The template isdocs/sdlc/plans/TEMPLATE.mdwhen the project overrides it, else$AI_HOME/skills/project-init/templates/plan.md. If the spec's Flagged concerns holds unresolved items, list them and have the human resolve or explicitly accept each one before planning.- If
superpowers:writing-plansis available, use it for the step granularity. Otherwise map each requirement to the files that must change (grep -n;Explorewhen a file is not found in a few calls), order the steps so each leaves the repository working, one step about one commit, and name the command or test that proves it. - Challenge it before writing: what could this break (callers, migrations, config, other environments) and which step is the riskiest — can it move earlier or be split. The answers go into Risks; mark the risky step.
- Copy the template to
docs/sdlc/plans/<slug>.md, fill Files that change / Order of work / Risks / Proof / Rollback; Proof maps every requirement R<n> to a test or command. One round of corrections; suggest the commitplan: <title>. Each row of the plan is later one task, planned below — unless the whole plan runs as one task through/rr-build-all(the project'sbatch.defaultinweight.json, or the developer's call): then every row becomes one step of that task, so give each row its files and its Consumes line (the rows it depends on) — that is what/rr-build-allreads.
A task → weight, mode, plan, start
1. Classify the request: feature, bugfix, refactoring, hotfix or
investigation — say which and why. using-riskrails.md in the shared tree
("$AI_HOME/riskrails/ai-path" using-riskrails.md) and
workflows.md#<kind> say what each kind requires. A refactoring and a feature
are never one task.
2. Name the files from the request, confirmed with grep -n. Recall what
the record holds about them — $STATE recall --files "<a>,<b>": open risks,
residual findings and earlier decisions are evidence for the weight and the
plan. Then weigh:
$STATE weight --files "<a>,<b>" --goal "<one sentence>" \
--declare behaviour=<0-2>:"<evidence>" --declare novelty=<0-2>:"<evidence>"
Seven lines: six axes, W and the mode. Scope, reversibility, coverage and the
hints are measured; behaviour, domain and novelty are yours to declare, each
with its evidence — a declaration without one is refused. Declare what the
change does to the project's decisions, not how long it feels. An unknown axis
scores 1. A hint raises an axis; a declaration below a hint does not lower it.
The table is .ai/policies/weight.json (weight.md mirrors it).
3. The mode is the weight's, not yours: raising it is free
(--mode <m> --by "<name>" on start); lowering it is the human's, from their
own terminal, or the command they picked in the dialog. Shape the plan by it:
| Mode | Plan | Who sees it before the first edit |
|---|---|---|
| light | inline: files, the test, the rollback, in one to five lines | nobody — go |
| guarded | plan mode: numbered steps, each with its files and tests | the human accepts plan mode |
| full | written: steps, files, tests, what stays the same, risks, rollback; architect on STRONG when the design is open |
the human reads and accepts it |
In every mode: a bugfix's first step is its failing test (C2); a change to
legacy behaviour without a test starts with a characterization test; the
rollback is one line, usually git revert. A plan over eight steps is advised
against, not refused — fewer, larger steps cost fewer handoffs.
4. Questions, now. Everything the plan depends on is asked before start,
so /rr-build stops only on a real blocker. Ask in the conversation, or on the
record with $STATE ask --topic <slug> … and $STATE answer --topic <slug> …
(see /rr-spec); a plan with an open (blocking) question is not accepted.
5. Start. One call registers the task, its weight and mode, its files and steps, and arms the scope guard:
$STATE start --goal "<one sentence>" --workflow <kind> --files "<a>,<b>" \
--tests "<t>" [--declare …] [--plan-file /path/steps.json] [--mode <m> --by "<name>"]
--plan-file is a JSON list [{"id": "s1", "files": [...], "tests": [...]}]
for a guarded or full plan with steps; a light task is one scope and needs none.
A step may add "after": ["s0"], the steps it waits for — /rr-build-all runs
the rest in parallel — and --batch --by "<name>" (or --no-batch) sets that
loop for the task; the default is batch.default in weight.json.
--files lists every file of every step — the guard refuses an edit outside
them. --tests names the task's own tests; a bugfix names its failing test. At
guarded and full, run start only after the human accepted the plan.
Then hand over: /rr-build continues from the state, in this session or the
next.
04 同仓库更多 Skills
sylius-fixtures
Load the Sylius fixture suite into the local database
deploy-staging
Deploy the current branch to staging and smoke-test the checkout. Use on "deploy to staging".
rr-status
暂无简介
rr-spec
Turn a rough idea into one file, docs/sdlc/specs/<slug>.md — an `## Intent` section (problem, outcome, affected users and systems, constraints, decisions taken) from a short one-question-at-a-time brainstorm recorded through state.py, then requirements and design from the architect agent on the STRONG tier, checked against the project's constitution, instruction files, skills and ADRs, ending with explicitly flagged concerns (especially contradicting policies). Accepts an existing docs/sdlc/intent/<slug>.md instead of the brainstorm. Use when starting a feature, "write the spec", "let's define what we want", "design this", before any plan or code.
rr-ship
Close the task — state.py close measures the two facts (a green whole-suite verify on the tree as it is now; every changed file inside the task's files), applies the mode table and either archives the task or, in full mode, requests the human's approval and prints the exact approve command for their own terminal. Writes the commit message body and suggests the commit; never merges or deploys, and commits, pushes or approves only as the command the human picked in the dialog, verbatim, once. Use on "/rr-ship", "close the task", "ship it", "ready for approval", after /rr-build or /rr-review.
rr-setup
暂无简介