Product Documentation / Deep Wiki
Transform the product that actually exists into accurate, navigable, maintainable documentation.
Core principle: evidence before documentation. Never promote an old README, issue, roadmap item, comment, or assumption into a current product capability without corroborating evidence.
Workflow
- Discover repository instructions, stack, modules, product surfaces, existing docs, tests, contracts, configuration and runtime entry points.
- Inventory meaningful capabilities before writing. Record status, implementation location, entry point, configuration, permissions, limitations, tests and evidence.
- Verify claims using the evidence hierarchy in
references/evidence-and-discovery.md. Uncertain capability = unverified, not stable.
- Design docs with progressive disclosure: TL;DR → README → Quick Start → guides → reference → architecture/internals.
- Write only useful surfaces. Do not create empty directories or boilerplate pages.
- Request visual evidence only where it improves understanding; delegate capture/automation instead of building a browser runtime here.
- Validate links, commands, examples, assets, diagrams and docs build with existing project tooling.
- Report coverage, drift, verification evidence and remaining gaps.
Load References Selectively
- Always read
references/evidence-and-discovery.md.
- README, Quick Start, guides, API/CLI/configuration, operations and troubleshooting: read
references/documentation-surfaces.md.
- Screenshots/GIF/video/Mermaid, documentation drift, coverage and final validation: read
references/visuals-drift-validation.md.
Responsibility Boundary
Own documentation discovery, information architecture, README, TL;DR, Quick Start, user/developer/operations guides, API/CLI/config references, troubleshooting, FAQ/glossary, What's New, coverage and drift detection.
Do not own feature implementation, bug fixing, global QA, release approval, Git workflow, production-readiness decisions, browser automation, or a parallel evidence store.
Use existing Maestro capabilities when evidence requires them:
skill-browser-agent for navigating/capturing UI evidence;
skill-webapp-testing for executable browser journeys;
skill-watch-evidence for existing video/screen-recording evidence;
skill-doublecheck for contested factual claims;
skill-verification-before-completion before claiming documentation is complete;
skill-adr only for an actual architecture decision record.
Guardrails
- Never invent features, metrics, APIs, config, screenshots or supported workflows.
- Never hide known limitations.
- Never create a docs-specific registry, runtime, browser engine or global evidence store.
- Never silently fix unrelated product defects found while documenting; report/escalate them.
- Never turn README into the entire manual.
- Never declare the product or release production-ready from documentation work alone.
Completion Evidence
A completion claim requires, when applicable: capability inventory, evidence-backed claims, accurate README/TL;DR, usable Quick Start, relevant feature/reference docs, reported visual artifacts, drift report, coverage report, and fresh validation of commands/links/builds that were actually executed.
1---2name: skill-deep-wiki3description: Use when documenting or auditing a software product or repository, creating or improving README, Quick Start, guides, API/CLI/configuration references, troubleshooting, visual documentation plans, or checking documentation drift after product changes.4category: documentation5risk: low6source: orquestrador-native7metadata:8 source: https://skillmd.com/skills/iapro-community/skill-deep-wiki9---1011# Product Documentation / Deep Wiki1213Transform the product that actually exists into accurate, navigable, maintainable documentation.1415**Core principle:** evidence before documentation. Never promote an old README, issue, roadmap item, comment, or assumption into a current product capability without corroborating evidence.1617## Workflow18191. **Discover** repository instructions, stack, modules, product surfaces, existing docs, tests, contracts, configuration and runtime entry points.202. **Inventory** meaningful capabilities before writing. Record status, implementation location, entry point, configuration, permissions, limitations, tests and evidence.213. **Verify** claims using the evidence hierarchy in `references/evidence-and-discovery.md`. Uncertain capability = `unverified`, not stable.224. **Design docs** with progressive disclosure: TL;DR → README → Quick Start → guides → reference → architecture/internals.235. **Write only useful surfaces.** Do not create empty directories or boilerplate pages.246. **Request visual evidence** only where it improves understanding; delegate capture/automation instead of building a browser runtime here.257. **Validate** links, commands, examples, assets, diagrams and docs build with existing project tooling.268. **Report** coverage, drift, verification evidence and remaining gaps.2728## Load References Selectively2930- Always read `references/evidence-and-discovery.md`.31- README, Quick Start, guides, API/CLI/configuration, operations and troubleshooting: read `references/documentation-surfaces.md`.32- Screenshots/GIF/video/Mermaid, documentation drift, coverage and final validation: read `references/visuals-drift-validation.md`.3334## Responsibility Boundary3536Own documentation discovery, information architecture, README, TL;DR, Quick Start, user/developer/operations guides, API/CLI/config references, troubleshooting, FAQ/glossary, What's New, coverage and drift detection.3738Do **not** own feature implementation, bug fixing, global QA, release approval, Git workflow, production-readiness decisions, browser automation, or a parallel evidence store.3940Use existing Maestro capabilities when evidence requires them:4142- `skill-browser-agent` for navigating/capturing UI evidence;43- `skill-webapp-testing` for executable browser journeys;44- `skill-watch-evidence` for existing video/screen-recording evidence;45- `skill-doublecheck` for contested factual claims;46- `skill-verification-before-completion` before claiming documentation is complete;47- `skill-adr` only for an actual architecture decision record.4849## Guardrails5051- Never invent features, metrics, APIs, config, screenshots or supported workflows.52- Never hide known limitations.53- Never create a docs-specific registry, runtime, browser engine or global evidence store.54- Never silently fix unrelated product defects found while documenting; report/escalate them.55- Never turn README into the entire manual.56- Never declare the product or release production-ready from documentation work alone.5758## Completion Evidence5960A completion claim requires, when applicable: capability inventory, evidence-backed claims, accurate README/TL;DR, usable Quick Start, relevant feature/reference docs, reported visual artifacts, drift report, coverage report, and fresh validation of commands/links/builds that were actually executed.