doc-fill

name	doc-fill
description	Fill one specific documentation gap directly, bypassing the orchestrator. Takes an `<area>` argument (e.g. `prd:auth`, `adr:adopt-postgres`, `module:billing-service`, `runtime:checkout-flow`, `slos`, `glossary`) and dispatches the appropriate specialist agent. Useful when you know exactly what you want next. Does not run the auditor first.
argument-hint	<area> e.g. prd:auth | adr:replace-mysql | module:billing-service | runtime:login | slos | risks | glossary
disable-model-invocation	false

/doc-fill — fill one specific gap

Use this skill when you already know what artefact you want to write and don't need the orchestrator's full audit + interview loop.

Argument syntax

/doc-fill <type>:<id>
/doc-fill <bare-type>          (for types that are singletons)

<area>	Dispatches	Writes to
intro	discovery-agent	docs/arc42/01-introduction-and-goals/overview.md
context	discovery-agent	docs/arc42/03-context-and-scope/{business,system}-context.md
prd:<feature>	discovery-agent	docs/arc42/03-context-and-scope/prds/<feature>.md
usecase:<name>	discovery-agent	docs/arc42/03-context-and-scope/use-cases/<name>.md
strategy	architect-agent	docs/arc42/04-solution-strategy/strategy.md
building-blocks	architect-agent	docs/arc42/05-building-blocks/overview.md + docs/diagrams/c4/workspace.dsl
adr:<title-kebab>	adr-writer	docs/arc42/09-decisions/NNNN-<title>.md
module:<name>	spec-writer	docs/detailed-design/<name>.md
runtime:<scenario>	spec-writer	docs/arc42/06-runtime/NN-<scenario>.md
crosscutting:<concern>	architect-agent	docs/arc42/08-crosscutting/<concern>.md
deployment	architect-agent	docs/arc42/07-deployment/deployment.md
slos	architect-agent	docs/arc42/10-quality/slos.md
risks	architect-agent	docs/arc42/11-risks/risk-register.md
glossary	architect-agent	docs/arc42/12-glossary/glossary.md

When to use

• You have a specific gap in mind: "write the auth PRD"
• You're filling in after the orchestrator stalled: e.g. orchestrator said INSUFFICIENT BRIEF for a particular file and you have the answers now
• You're maintaining: a new feature lands → you want one new ADR + one new module spec, no ceremony
• You want fast turnaround: skip the audit overhead

When NOT to use

• You don't know what's missing → use /doc-status first
• You want guided discovery from zero → use /doc-init instead
• You want the orchestrator to keep state between dispatches → use /doc-init (this skill is a one-shot)

Behaviour

1. The skill parses <area> and identifies the target agent
2. The user is prompted (one focused turn) to provide the minimum brief the specialist needs (e.g. for adr:adopt-postgres: drivers, options, chosen rationale)
3. The skill dispatches the specialist with that brief
4. The specialist writes the file and returns its standard structured summary
5. The skill suggests running /doc-status to verify the gap closed

Examples

/doc-fill prd:authentication
  → discovery-agent asks: "What is the user story for auth, in one sentence?
    What 3 functional requirements are MVP? What is the availability NFR?"
  → writes docs/arc42/03-context-and-scope/prds/authentication.md
  → suggests: "/doc-status to verify; or /doc-fill usecase:login next"

/doc-fill adr:choose-event-broker
  → adr-writer asks: "What are the 3–5 drivers? What 2–3 options were
    considered? Which won and why? Y-statement?"
  → writes docs/arc42/09-decisions/0007-choose-event-broker.md
  → suggests: "/doc-status to verify; remember to update strategy.md"

/doc-fill module:notification-service
  → spec-writer asks: "What does this module expose (API methods)?
    What data does it own? What are its 2 worst failure modes? What's
    one rejected alternative design?"
  → writes docs/detailed-design/notification-service.md
  → suggests: "/doc-fill runtime:notification-delivery if you want a
    sequence diagram for the main flow"

Implementation note

This skill is a router. It dispatches to existing specialist agents defined in .claude/agents/. Edit those agents to change writing behaviour; this skill only handles dispatch routing.

---

See also

• /doc-init — full orchestrated conversation
• /doc-status — coverage check
• .claude/agents/README.md — agent inventory + rubric