Thin orchestrator for decision-grade approach selection. Converts evidence-first fact-finds into a chosen approach, rejected alternatives, and a planning handoff artifact ready for /lp-do-plan.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.
Instruções da origem · Visualização somente leitura
name
lp-do-analysis
description
Thin orchestrator for decision-grade approach selection. Converts evidence-first fact-finds into a chosen approach, rejected alternatives, and a planning handoff artifact ready for /lp-do-plan.
Analysis Orchestrator
/lp-do-analysis is the convergence layer between fact-finding and planning.
It owns:
Discovery/intake/dedupe
Analysis gates
Track routing to specialized analysis modules
Analysis artifact persistence using external templates
Mandatory critique via /lp-do-critique
Optional handoff to /lp-do-plan
It does not decompose executable tasks. That is /lp-do-plan's job.
Global Invariants
Operating mode
ANALYSIS ONLY
Allowed actions
Read/search files and docs.
Run non-destructive validation commands needed for evidence.
Edit analysis docs.
Prohibited actions
Production implementation changes, refactors, migrations, or deploy operations.
Destructive shell/git commands.
Task decomposition beyond planning handoff notes.
Re-opening broad discovery that belongs in /lp-do-fact-find unless the fact-find is insufficient.
Inputs Priority
Fact-find handoff packet (docs/plans/<slug>/fact-find.packet.json) when present
Fact-find brief (docs/plans/<slug>/fact-find.md)
Existing analysis (docs/plans/<slug>/analysis.md)
User request/topic/card ID
Current repository evidence
Use the packet-first load order from . Read the full only when the packet is insufficient or a path-specific claim needs verification.
Carry forward the outcome contract from the fact-find without fabricating operator-authored rationale.
For Execution-Track: code | mixed, fill ## Engineering Coverage Comparison using ../_shared/engineering-coverage-matrix.md.
Phase 5.5: End-State Operating Model (Non-omittable)
Before persist, write ## End-State Operating Model.
This section may be a single line None: no material process topology change only when the chosen approach does not alter any multi-step process, workflow, lifecycle state, CI/deploy/release lane, approval path, or operator runbook.
For process-affecting work, describe the target state area by area:
current state
trigger/start condition
step-by-step end-state flow after the chosen approach lands
what remains unchanged
unresolved seams/risks that planning must absorb
If the target process cannot be explained without hiding behind task-level detail, the analysis is not decision-grade yet.
Phase 6: Persist Analysis
Write/update:
docs/plans/<feature-slug>/analysis.md
canonical sidecar after validators pass: docs/plans/<feature-slug>/analysis.packet.json
Status policy:
Default Status: Draft.
Set Status: Ready-for-planning only when:
mode is analysis+auto (default) or user explicitly wants planning handoff now, and
analysis gates pass, and
no unresolved blocker remains in the chosen approach
Phase 7: Critique Loop (1–3 rounds, mandatory)
After the analysis is persisted (Phase 6), run the critique loop in analysis mode.
Load and follow: ../_shared/critique-loop-protocol.md
Record once per materially updated analysis artifact.
Include the analysis module(s) actually loaded plus the upstream packet and any full upstream artifact paths that were materially loaded as context.
Include scripts/validate-analysis.sh on every analysis telemetry record, plus scripts/validate-engineering-coverage.sh when required by track.
Codex token usage is auto-captured when CODEX_THREAD_ID is available.
Claude token usage is auto-captured via project session logs (sessions-index.json → debug/latest fallback). Explicit --claude-session-id still takes priority when supplied.
Phase 8: Optional Handoff to Plan
If eligible and mode is analysis+auto, invoke /lp-do-plan <feature-slug>.
