docs-onboarding

Onboard an existing project into the scaffolded Harness wiki by inspecting code, docs and backlog, closing reconstruction decisions, then handing accepted flows to requirements-phase for architecture/

SKILL.md

GitHub

Docs Onboarding

Use this skill to turn an existing repository into initial Harness wiki context after hi init.

Workflow

Before naming or rewriting domain concepts, read the canonical glossary in the routed requirements-grill status artifact. Preserve its canonical terms; route proposed terminology changes through requirements-grill instead of silently renaming them.

  1. Read repo guidance and the existing project-owned wiki guidance.
  2. Read any existing backlog provider and wiki framework decisions. Resolve their locations from project guidance; hi init does not generate those files.
  3. Discover backlog context:
    • Probe the pinned remote backlog provider first when connector/tool/auth access exists.
    • Inspect local backlog clues after that: existing specs, issue matrices, backlog markdown, project docs, and .devpunks metadata.
    • Treat remote backlog evidence and local clues as equally important and mandatory when available.
    • If either lane is missing or inaccessible, ask the developer for the backlog location or access path before assuming no backlog exists.
  4. Build and persist the Project Map before asking developer questions.
  5. Run a targeted requirements-grill against the Project Map for project-wide reconstruction decisions.
  6. Activate write-backlog's backlog-initialization branch to reconcile the Product/Backlog Root, Product Areas, Initiatives, V* context, and saved views from accepted wiki and provider evidence.
  7. Update onboarding navigation and indexes using references/artifact-contract.md.
  8. Hand every candidate existing flow accepted by the onboarding grill to $requirements-phase as an explicit human handoff, then stop. For each flow, name the exact $requirements-phase command and a brief carrying the confirmed onboarding grill status/log, its open per-flow decisions, its Project Map evidence, and its onboarding projection boundary. Requirements Phase compiles create-architecture before create-spec. The boundary lets an agent-ready spec and its mapped architecture reach write-backlog only when reconstructed evidence already supplies the exact existing Epic/Story and accepted projection authority required by that branch. Otherwise record the missing authority and exact human Finder invocation in the brief; the compiled pair is retained and the backlog delta returns blocked without provider mutation. Onboarding never invokes a Finder implicitly and does not create higher-level placement.

Discovery Depth

Inspect entrypoints, routes/pages, commands, API handlers, schemas/models, tests, config/env examples, and existing docs.

Do not deep-read every implementation file by default. Deep-read implementation files only when needed to explain a flow, resolve a contradiction, or map a data boundary.

Targeted Grill

Grill contradictions, missing backlog location, unclear feature ownership and inferred requirements before either compiler runs. Observed code-only facts may remain Inferred in the Project Map as evidence; they become accepted requirements or design choices only through confirmed intent and the recorded grill decisions. Carry unresolved per-flow material decisions in the step 8 handoff brief; an inferred current implementation is not permission to choose architecture or requirements.

Resumability

Persist the Project Map and grill status before developer questions. On resume, continue from saved discovery/grill state instead of rescanning from zero.

Guardrails

  • Code can prove observed behavior, not original intent.
  • Reconstructed spec history must be evidence-backed.
  • Do not invent implementation notes, rationale, or tech-debt history without supporting artifacts.
  • Use only write-backlog for physical provider mutation.