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/dev/patterns/satellite-review.md
View on GitHub# Worker + Satellite Review Pattern
After completing any non-trivial implementation step, the worker agent itself invokes `@architect` as a satellite reviewer before presenting verification to the developer.
Apply `{{provider:skills_root}}/dev/runtime/delegation.md`. If nested delegation is unavailable,
the orchestrator dispatches the independent reviewer after the worker returns.
## When to trigger
Trigger a satellite review after any step that:
- Adds or changes infrastructure definitions, IAM policy, or cloud resources
- Introduces a new API endpoint, module, or cross-layer integration
- Changes data flow between containers
- Involves a workaround or non-obvious implementation decision
Skip for mechanical steps: rename, dependency bump, config value with no structural impact.
## Review protocol
1. Worker completes the step
2. Worker invokes `@architect` with:
> "Read `docs/Target Architecture (HLD)/README.md` to find the active version. Review the implementation of Step N from plan `<1-plan.md path>`. Check: HLD container boundaries and data-flow direction, layer best practices (infrastructure: naming/security rules/IAM least-privilege; backend: p95/N+1; frontend: accessibility), NFRs located through the active architecture documentation index. Do not write or edit files; return findings only. Verdict: **go** or **reiterate** — one finding per line."
3. If **go**: worker hands off to the orchestrator (never directly to the developer)
— see Handoff below.
4. If **reiterate**: return findings to the orchestrator for `patterns/review-gate.md` recording before fixes. The orchestrator owns the report, delegates in-scope fixes, and requests only unresolved human decisions. Repeat relevant review after fixes.
5. **Maximum 3 iterations.** If still unresolved after 3, worker escalates to the orchestrator with a summary of the stalemate, for the orchestrator to present to the developer.
## Handoff — report to the orchestrator
A satellite `go`, and the worker's own account of what it did, are both inputs to
the orchestrator's step checkpoint (`references/04-implement.md` Step 3). Neither
replaces it.
1. Worker hands the orchestrator a structured summary: what changed (files
touched, brief rationale), what the worker itself verified and the real result,
and the satellite's verdict. This is a report, not a completed step. The
orchestrator checks the change actually landed (`git status`/`git diff`, real
command output) rather than trusting the self-report.
2. The orchestrator then runs the step checkpoint itself: run the plan's
verification items, auto-approve when they qualify, or stop and ask the
developer when they don't. A worker never talks to the developer directly.
## Why
Without a review loop, best-practice violations (hardcoded names, wrong security rules, removal policies, layer boundary breaches) surface only at PR review time — after deployment, requiring a dedicated refactor step. Without an explicit handoff step, a satellite `go` or a worker's own report can get silently treated as done, skipping the orchestrator's own check of what actually landed.
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: 471k | 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 |