Mirror Neuron Documents

Documentation Style

Author and review implementation-backed MirrorNeuron documentation.

Documentation Style

Use this guide when you create or substantially update a MirrorNeuron documentation page. Documentation is part of the product: it must help a named reader reach a safe, observable outcome without guessing.

Write a brief before drafting

Record the following in the issue, pull request, or page draft:

ItemWhat to define
ReaderEvaluator, application developer, blueprint author, operator, integrator, contributor, or maintainer.
OutcomeOne observable result the reader can achieve.
Page typeTutorial, how-to, reference, explanation, troubleshooting, migration, or decision record.
ScopeThe exact command, API, component, workflow, or behavior covered.
Out of scopeNearby topics intentionally excluded.
Sources of truthCode, CLI help, tests, schemas, generated API definitions, or release artifacts used to verify claims.
ValidationThe documentation build and the smallest relevant product command or test.

Keep page types distinct

TypeReader needRequired content
TutorialLearn a complete first experience.Prerequisites, ordered steps, verification, cleanup, and next steps.
How-toComplete one operational task.Assumptions, procedure, verification, rollback, and symptom-led help.
ReferenceLook up exact facts.Declared scope, exact syntax or fields, defaults, side effects, errors, and small examples.
ExplanationUnderstand a design choice.Problem, mental model, boundaries, guarantees, non-guarantees, failures, and tradeoffs.
TroubleshootingRecover from an observable failure.Symptom, likely causes, read-only diagnostics, resolution, verification, and escalation evidence.

Do not turn a reference page into a tutorial or hide operational procedures inside architecture prose. Link to the canonical page instead.

Commands and examples

  • Use fenced bash, json, python, text, or other appropriate code blocks.
  • State the working directory when it matters.
  • Use only commands verified against CLI help, tests, schemas, or a representative runtime.
  • Use a checked-in example only when its requirements and safety boundaries are documented.
  • Use placeholders such as <job_id> and define them immediately after the block.
  • Show a stable success marker only when it is verified. Otherwise tell the reader which command exit status or observable state proves success.
  • Include cleanup for jobs, services, ports, generated files, or persistent state created by the procedure.

Accuracy and safety

  • Use the implementation as the source of truth. Never infer a command, default, port, guarantee, or compatibility policy.
  • Keep one canonical page for durable facts such as CLI syntax, API contracts, environment variables, and manifest fields.
  • State what a feature does not guarantee, especially for retries, external side effects, local execution, sandboxing, privacy, and clustering.
  • Warn before actions that delete data, expose listeners, pass secrets into worker code, execute unreviewed bundles, or join a cluster trust domain.
  • Do not publish agent instructions, release procedures, templates, obsolete diagrams, or references to removed components as product documentation.

Review checklist

  • The title, reader, outcome, and primary page type are clear in the opening screen.
  • Important claims cite or were checked against an authoritative implementation source.
  • Commands, links, examples, and stated defaults were validated.
  • Each state-changing procedure includes verification and applicable cleanup or rollback.
  • Security, data movement, secrets, network exposure, and non-guarantees are visible where relevant.
  • Related pages point to the next useful action.

On this page