router
Find where a request stands on the line when the next step is unclear: resume an open or stopped change at the step its "next" line names, choose the start the request's situation calls for, or send a new change to its proposal. Use for continuity ("what were we doing?"), for a request no start fits, and for doubt about the next step.
- License 许可
- MIT
- 最近更新
- 2026-10-21
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- router
- description
- Find where a request stands on the line when the next step is unclear: resume an open or stopped change at the step its "next" line names, choose the start the request's situation calls for, or send a new change to its proposal. Use for continuity ("what were we doing?"), for a request no start fits, and for doubt about the next step.
02 资源文件 2
- 参考文档references/greenline.md721 B
- 参考文档ROUTING.md4.4 KB
03 SKILL.md 正文
Router
AGENTS.md owns the request loop, the modes and the roster by kind. This start finds where a request stands: an open or stopped change to resume, a start to choose, or a new change to propose. The ledger guide owns the change record's format.
Greenline's paths and commands for this skill stand on its references page, references/greenline.md: read it when the work lands in the workspace.
Resume a change
Read the open and stopped change records in the ledger, the
exact references in the request and the relevant files; the workspace's status
lists each change with its state and its "next" line. A change stopped on
another branch, local or remote-tracking, is found by its record's commit,
which every stop makes, and read from that branch. Read the candidates before deciding which change the
owner means: recency alone is not ownership. A record's "next" line names
the step reached and the action that resumes it. Resume there on the
change's branch, the one its record's commit sits on, switching to it
first when the session sits on another, and making a local branch from a
remote-tracking one; go on under the record's mode, with
no new proposal, and set the record's state open as your first write in
it. Ask only when several changes remain plausible, and say what each would
produce.
Choose the start
When this session holds a change and the request asks for another
change, stop that change first, as the ledger guide says. A
session that resumes, or is asked "what were we doing?", with no open or
stopped change on any branch says there is nothing to resume, proposes
nothing and asks the owner for the request. When no open or stopped change
is the request's, load the one start whose row in AGENTS.md names the
situation the request begins in, or at an orientation the start AGENTS.md
names for it, and no other; a start the owner names runs. architecture-survey loads at an orientation, a repository with
documents and no change record on any branch, local or remote-tracking, or
when the request asks for a survey or the owner names it; where a change
record exists, it is never an orientation. A start that did not run is
never named in the record. A request that asks for a change and fits no
start goes straight to its proposal: the mode, then the line from frame.
A step asked for by name runs alone. ROUTING.md, the map of the line,
separates two methods that could produce different work.
Watch the habits
While any step runs, watch for the situations the habits' rows in
AGENTS.md name. Load a habit only when the situation its row in AGENTS.md
names arises in the work at hand, never ahead of it or beside another habit
in case it might. secrets loads only where a secret is in reach: a key, a
token, a credential or a .env file that the work reads, writes, stages in
a commit or builds into an image, or a credential that may have leaked; a commit or
an image with none in reach loads nothing. ponytail loads only in its own
rows: before writing code the request leaves open, a refactor or a
simplification of code that works, a new internal API replacing an old one
that still has callers, or a question about what was deferred; writing code
the request plainly asks for is none of them. A habit whose output is a
durable file the owner did not ask for is offered in one line and made on
the owner's yes.
Handoff
Consumes: the request and the open or stopped change records under .greenline/ledger/ Produces: the record's state in .greenline/ledger/<id>.md, open for the change resumed and stopped, with its "next" line and committed, for the change set aside; otherwise nothing durable: the start loaded, the proposal made, or, with nothing to resume, the owner asked for the request, in the reply Next: the step the resumed change's "next" line names, the start the situation calls for, or frame under the mode the owner approves
04 同仓库更多 Skills
use-garden
Consulting garden, the separately installed guidance command this repository's owner enabled, when an open engineering choice the repository does not settle, made in a change or only recommended, would be informed by published guidance. Use for a choice between approaches, a pattern or pitfall for a technology or concern the work touches, or a unit a repository decision names; not for a question about this repository's own facts or a change whose choices are already settled.
use-garden
Consulting garden, the separately installed guidance command this repository's owner enabled, when an open engineering choice the repository does not settle, made in a change or only recommended, would be informed by published guidance. Use for a choice between approaches, a pattern or pitfall for a technology or concern the work touches, or a unit a repository decision names; not for a question about this repository's own facts or a change whose choices are already settled.
what-to-build
Propose what this product could do next at three altitudes: sharpen a shipped slice, extend one, or open ground the product does not stand on yet. Use for 'what should we build next', 'any ideas for this app', 'what's missing', 'is this done or is there more in it', and quick-win hunts. Not for refactors, which belong to the architecture survey.
wayfinder
Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets in a plan directory's map.md, and resolve them one at a time until the way to the destination is clear.
verify
Prove a change's claims with fresh evidence: restate each item of its acceptance so it can fail, capture a baseline and a treatment with the same command, compare the raw artifacts, and return VERIFIED, NOT VERIFIED or INCONCLUSIVE for each, at the depth the change's mode gives the verify step. Use when the mode reaches verify, or when the owner says "verify this", "prove it works" or "did this fix it".
tdd
Test-first development at any step: discover the repository's own test commands, name the break a test catches, watch it fail for the intended reason, make it pass with the smallest change, and add a new test only where it catches a realistic failure the existing ones would miss. Use when writing or changing a test, building a feature or fixing a bug test-first, or deciding whether a change needs a new test; also on 'red-green-refactor' and integration tests.