Mapear

Translation note: Originally authored in Portuguese (pt-BR) by Igor Uehara (igoruehara/spec-driven, MIT). Translated to English by this hub to keep the repository language consistent. Original content unchanged in meaning; see the upstream repo for the pt-BR source.

Skill: Map the current state (as-is)

Produces a portrait of a project already in progress. First read the code, then ask only what the code does not reveal. It is idempotent: re-running updates docs/architecture/assessment.md.

Process

1. Automated mapping: identify stack, folder structure, architecture style, couplings, tests/CI, logs/metrics/tracing. Infer the implicit bounded contexts from the code organization. In large repos, delegate the scan to an exploration subagent (see docs/engineering/_templates/subagent.template.md) to keep the context lean.
2. External inputs (if any): if /integracoes connected GitHub/cloud/Confluence (validated account), use them to enrich the as-is. Cite the source.
3. Gap interview (AskUserQuestion): current business intent and North Star; biggest pains/risks today; domain terms that confuse the team; what MUST NOT break; team context and size.
4. Gap analysis: compare the as-is with the SDD standard across the 5 axes (tech stack, architecture, infrastructure, quality, observability). Mark risk (low/medium/high).
5. Historical decisions: list structural choices already made without a record → they become retroactive ADRs (status: accepted, recording the historical rationale).

Outputs

• docs/architecture/assessment.md (use docs/architecture/_templates/assessment.template.md).
• List of retroactive ADRs to create in docs/architecture/adr/.

Next step

• Inside /kickoff: feed the assessment + gaps back into the 5 axes and the roadmap.
• Standalone: suggest /roadmap (to prioritize the mapped debts) or /camada-agentica.