findings-routing
Author the dispatch-plan payload a flywheel findings-routing session builds over a bolt's queued findings — the whole findings inbox seeded for one operator…
Skill by agentplot
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/agentplot-findings-routing
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/findings-routing
What's inside
- Skill name
findings-routing- Description
- Author the dispatch-plan payload a flywheel findings-routing session builds over a bolt's queued findings — the whole findings inbox seeded for one operator round (dispatch runs it) into the current bolt, a successor bolt, the source intents, or drop. Use whenever a work order charges a findings-routing run on a bolt's queue.
- License
- MIT
- Instructions
- 84 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Flywheel findings-routing — the round over a bolt's queue
You are a one-shot session charged by the bolt loop when it stops at a
non-empty queue. Test and code-review sessions wrote each finding as a
`state:queued` item at birth — the bolt's **findings inbox**, durable
but inert. Your whole job is one dispatch plan over that inbox: the
shared protocol at `skills/_reference/dispatch-plan.md` is the practice,
and this skill only says what its construction origin pins down.
## The rows are the queued items
Read every open `state:queued` non-container item on the bolt's
milestone, plus the finding comments they carry. Each becomes one row,
seeded with your judgment — like dispatch's intake issues, they pre-exist
as the inbox and are the subject of proposals: the apply moves, cards,
or closes them, and never duplicates one.
The containers you may propose, and what each route does at apply:
- **this bolt** — work its merge criteria need: a unit card on the open
milestone, in bolt-planning's exact grammar, citing the finding item;
the finding closes `closed:superseded` with the card as its successor
pointer, because the card is now the tracked object and expansion
births the work items.
- **a successor bolt** — related work this delivery does not need: a
new `bolt/<slug>` container, named per bolt-planning's rule (the
deliverable, never the task list), its cards filed the same way.
- **a source intent** — a finding that indicts the design rather than
the build: the item moves onto that intent's milestone with the
`type:*` of the session that will work it, and joins the elaboration
your plan composes there.
- **backlog** — stays queued on the bolt, explicitly out of this round.
- **drop** — closed `closed:declined`, with the reason as the closing
comment.
## The payload is real files on the bolt change
You run on the repo that owns the bolt change. Write the payload under
the bolt's own change directory —
openspec/changes/bolt-<slug>/sessions/<date>-findings-routing/close/
— the same file shapes the protocol names, and commit it by pathspec
before any tracker write, pushing what you commit: your directory is a
shared checkout, so the standing staging rule binds you — never `-a`,
never `add -A`.
## Publish — you run no round
You author and publish; dispatch assembles every standing payload into
the one round and applies the operator's word (the protocol's
"Publishing a payload"). Three acts, then settle:
1. Commit the `close/` payload by pathspec and **push your branch** —
dispatch has no checkout and reads the files through the contents
API at your commit's SHA.
2. Write the round-payload marker as a comment on the anchor item — the
bolt's unit parent, or its lowest queued finding — with
`ORIGIN: construction` and the pushed SHA.
3. Add the `dispatch:standing` label to the same item.
No lavish page, no digest, no apply, no Ready, and no `needs-operator`:
the payload stands until a round consumes it. The loop that charged you
is already stopped, so nothing blocks; an operator who never calls a
round loses nothing — the inbox waits, and a send-back comes back to a
fresh session as the note on the anchor. Applying the plan (dispatch's
act) drains the inbox the way applying a guard's plan empties the
guard: a later run against an unrouted queue proposes again; against an
empty one, the loop charges nothing. Nothing the plan proposes reaches
GitHub before the approval.
## What you never do
You never fix a finding, build anything, or edit the built repo — you
route. A finding that needs no routing decision does not exist: backlog
is a decision too. Machinery findings follow the finding-routing rule: your
report, never the tracker.
## What you report
One line per row with its seeded route, the payload's anchor item and
SHA, and anything you left out of the plan and why.From agentplot/flywheel, MIT.
More from agentplot/flywheel
- 0.524score
interactive
agentplot
Build the lavish page a flywheel interactive-design session works — one artifact carrying every decision the batch aims to close, with options, trade-offs…
SkillBFreshness: BClaude Code +9 - 0.524score
bolt-planning
agentplot
Run a bolt planning session — read a system's design book, the built repo's implemented specs, and the changes in flight; carve the remaining gap into one or…
SkillBFreshness: BClaude Code +9 - 0.524score
construction
agentplot
Run the flywheel construction loop — the bolt loop drives an approved plan's work items through spec, build, verify, merge and landing across built-repo bolt…
SkillBFreshness: BClaude Code +9 - 0.524score
fleet
agentplot
Drive an org's flywheel fleet from its manifest — flywheel up starts the server that runs the loops, flywheel dispatch starts the org's standing dispatch…
SkillBFreshness: BClaude Code +9 - 0.524score
import
agentplot
Run the messy-repo import a flywheel import-charged design session works — inventory a corpus of planning docs, cut it against the destination books, convert…
SkillBFreshness: BClaude Code +9 - 0.524score
inception
agentplot
Run the flywheel design loop — dispatch raw ideas into intent changes and tracker items, run design sessions that close decisions and write the books and the…
SkillBFreshness: BClaude Code +9 - 0.521score
build
agentplot
Apply a specced work item the way a flywheel build construction session does — /opsx:apply on a nested construction worktree, several items per session…
SkillBFreshness: BClaude Code +9 - 0.521score
code-review
agentplot
Run the read a flywheel code-review construction session takes over built work — adversarial against the spec and its cited decisions, or through a persona…
SkillBFreshness: BClaude Code +9 - 0.521score
human-code-review
agentplot
Run the operator's review round a flywheel human-code-review construction session prepares — built diffs put in front of the operator through plannotator, the…
SkillBFreshness: BClaude Code +9 - 0.316score
openspec-propose
openspec
Propose a new change with all artifacts generated in one step.
SkillBFreshness: BClaude Code +9 - 0.316score
openspec-explore
openspec
Enter explore mode - a thinking partner for exploring ideas, investigating problems, and clarifying requirements.
SkillBFreshness: BClaude Code +9 - 0.316score
openspec-archive-change
openspec
Archive a completed change in the experimental workflow.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 38
- Last commit (this folder)
- Aug 25, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/agentplot/flywheel
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from agentplot
- 0.524score
interactive
agentplot
Build the lavish page a flywheel interactive-design session works — one artifact carrying every decision the batch aims to close, with options, trade-offs…
SkillBFreshness: BClaude Code +9 - 0.524score
import
agentplot
Run the messy-repo import a flywheel import-charged design session works — inventory a corpus of planning docs, cut it against the destination books, convert…
SkillBFreshness: BClaude Code +9 - 0.524score
inception
agentplot
Run the flywheel design loop — dispatch raw ideas into intent changes and tracker items, run design sessions that close decisions and write the books and the…
SkillBFreshness: BClaude Code +9 - 0.524score
bolt-planning
agentplot
Run a bolt planning session — read a system's design book, the built repo's implemented specs, and the changes in flight; carve the remaining gap into one or…
SkillBFreshness: BClaude Code +9 - 0.524score
construction
agentplot
Run the flywheel construction loop — the bolt loop drives an approved plan's work items through spec, build, verify, merge and landing across built-repo bolt…
SkillBFreshness: BClaude Code +9 - 0.524score
fleet
agentplot
Drive an org's flywheel fleet from its manifest — flywheel up starts the server that runs the loops, flywheel dispatch starts the org's standing dispatch…
SkillBFreshness: BClaude Code +9
Cite this page
findings-routing skill. Agent Mart. https://agentmart.dev/skills/agentplot-findings-routing (last updated 2026-10-24)