DataArt/artisyn-implementation-harness / skills/implementation-harness
implementation-harness
FlaggedInstall the portable AI-dev harness, rebuild its operational graph standalone, harvest project skill improvements into its canonical templates, package direct canonical edits into a portable bundle, or apply canonical updates to an existing project. Use /implementation-harness launch, /implementation-harness reindex-graph, or /implementation-harness patch harvest|prepare|apply.
canonical/skills/design/references/00-decision-router.md
View on GitHub# Decision router
Shared loop mechanics for a stage running against two or more confirmed
decisions (`design/00 Frame.md`'s `## Decisions` section, confirmed at
Frame's dedicated decisions sign-off — Step 2). `patterns/multi-decision-split.md`
owns which artifacts split per decision and which stay single; this file owns
how a stage actually runs the loop without multiplying its stops.
## The rule
A stage with N confirmed decisions still closes with exactly one
Sign-off → Report → Commit — never one per decision, and never ambiguous
about which steps repeat. Only the content-producing steps loop; the closing
steps run once, after every decision's content is drafted, presenting all N
together in that one sign-off.
## Per-stage loop boundary
| Stage | Loops (once per decision) | Closes once, not looped |
| -------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------- |
| Research | Steps 2, 2b, 3 (delegate → OQ triage → save) — produces `design/01 Research - {Decision}.md` each | Steps 4–6 (Sign-off, Report, Commit) |
| Evaluate | Steps 1b–6 (resolve viability OQs → matrix → recommendation → closing WAF → save) — reads the matching `design/01 Research - {Decision}.md`, produces `design/02 Evaluate - {Decision}.md` each | Steps 7–9 (Sign-off, Report, Commit) |
| Solution | Step 3 only (`solution/Design - {Decision}.md`) — `Architecture.md`, `Stack.md`, `Efforts.md` stay single files regardless of decision count | Steps 8–10 (Sign-off, Report, Commit) |
A stage with at most one decision ignores this file entirely: no looping, no
`{Decision}` suffix, proceed exactly as the steps read.
## Tracking each decision as a task
Apply `{{provider:skills_root}}/design/patterns/task-tracking.md` with the
confirmed decisions as the unit list: one task per decision, in the order
Frame listed them, as children of the stage-level task the router already
created for Research/Evaluate/Solution — the same shape `dev`'s Implement
stage uses for its plan steps. Mark each decision's task `in_progress` when
its loop iteration starts and `completed` only once that iteration's file is
actually written (not on a self-report). This is native-tool metadata for the
developer to watch progress by (a Claude task, a Codex todo, or whatever the
host provides) — `design/Report.md` stays the durable record regardless of
which native tool is present or absent.
## The closing sign-off, with N decisions
Present every decision's result together, named explicitly — not "see the
files," the actual per-decision summary (options count, recommendation, P1s,
whatever that stage's sign-off already shows for one decision, repeated once
per decision in the same message). One `wait start`, one acceptance check,
one commit covering every decision's files from that stage.
## Why a separate file
`patterns/multi-decision-split.md` is a content/file-split policy — which
artifacts become per-decision files. This is a control-flow rule — where the
loop starts and stops within a stage's own step sequence, so a stage with
multiple decisions never turns into multiple approval gates. Keeping it
separate means each stage reference states one line pointing here, instead
of restating — and inconsistently scoping, as "loop this phase" vs. "loop
this step" demonstrated — the same rule three times.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |