From deliverable
Drafts business requirements (BRD/PRD) section by section through structured intake, discovery research via sub-agents, role-based interviews, and approval gates. Produces brd.md, decisions.md, and open-questions.md. Use when the user asks for a BRD, PRD, "spec this feature," or "requirements for X."
How this skill is triggered — by the user, by Claude, or both
Slash command
/deliverable:deliverable-brdThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Load on demand from the sibling `deliverable/` skill folder: `roles/*.md`, `templates/*.md`, `sub-agents/*.md`, `references/*.md`.
Load on demand from the sibling deliverable/ skill folder: roles/*.md, templates/*.md, sub-agents/*.md, references/*.md.
Draft business requirements through structured interview and research. Works section by section with your approval at every step. Pushes back on vague answers. Spawns bounded sub-agents for competitive research and feasibility checks.
Announce at start: "I'm using the deliverable-brd skill to draft a BRD through structured phases with approval gates."
NEVER draft the entire BRD in one shot. NEVER write multiple sections in a single turn. NEVER advance to the next phase without explicit user approval. NEVER write to disk without announcing and getting confirmation.flowchart TD
P1[Intake] --> P2[Discovery Research]
P2 --> P3[Problem & Users]
P3 --> P4[Scope & Metrics]
P4 --> P5[Risks & Assumptions]
P5 --> P6[BRD Sign-off]
P6 --> P7[Dual-voice BRD Review]
P7 --> DONE[Done → suggest SRS]
Five questions, one at a time:
greenfield / feature / internal / autoprd-lite / exec-onepager / nonerfc / acceptance-tests / nonerisks-register / planning-handoff / roadmap / noneAuto-bump signals: If user mentions PII, payment data, PHI, regulated industry — note in state.md for deliverable-srs skill.
Propose sub-agents from sub-agents/:
competitor-scan — max 5 candidates, public info (~2 min, ~3k tokens)feasibility-check — library maturity, licensing (~2 min, ~2k tokens)prior-art — existing solutions, patterns (~2 min, ~2k tokens)User approves which to run. Parallel dispatch OK. Raw → cache, distilled → docs/requirements/research/.
Interview using roles/sponsor.md, roles/pm.md, roles/designer.md. Draft:
Interview using roles/pm.md, roles/tech-lead.md, roles/designer.md, roles/qa.md. Draft:
Interview using roles/pm.md, roles/tech-lead.md, roles/security-legal.md*. Draft:
[ASSUMPTION]*security-legal loaded if auto-bump signals detected
Present full BRD. List all [ASSUMPTION] and [OPEN] items. Ask for sign-off.
Dispatch sub-agents/dual-voice-reviewer.md for independent second opinion on scope and success metrics. Present findings. User decides what to address.
| Section | greenfield | feature | internal |
|---|---|---|---|
| §market-sizing | required | skip | skip |
| §competitive-landscape | required | optional | skip |
| §pricing-viability | required | optional | skip |
| §target-user | required | required | "adopter persona" |
| §success-metric | required (north-star) | required | required (adoption) |
Skipped sections omitted entirely — no empty stubs.
Orient → Work → Present → Approve/Edit/Revise/Kill → Commit.
User can't answer? Suggest the deliverable-interview skill to generate templates. Pause phase, resume when answers return.
[ASSUMPTION] and [OPEN] tags everywhere.When the user selected roadmap as a cross-cutting extra during Intake, generate it after BRD sign-off.
After drafting the roadmap content, ask:
"Ready to write the roadmap. What format would you like? 1. Markdown only (
roadmap.md) 2. Excel only (roadmap.xlsx) 3. Both"
Write docs/requirements/roadmap.md using templates/roadmap.md.
Gather the following extra sections using the four-beat rhythm — one at a time, present → approve → commit.
Ask: how many sprints does the roadmap cover? For each sprint: name (e.g. "Sprint 1") and date range (e.g. "01/05 - 14/05").
Ask: are there named releases tied to specific sprints? If yes, for each release: release name and which sprint it lands in.
For each feature or work item on the roadmap, ask which sprints it is in development (D) vs released (R). Use the Now/Next/Later items from the roadmap content as the feature list. For each:
D (development), R (release), or leave blankAfter all extra sections are approved, write docs/requirements/roadmap-data.json with this shape:
{
"project_name": "<name>",
"date": "<YYYY-MM-DD>",
"sprints": [{ "name": "Sprint 1", "dates": "01/05 - 14/05" }],
"releases": [{ "name": "Release 1", "sprint": "Sprint 2" }],
"features": [
{
"tool": "",
"element": "",
"sub_element": null,
"timeline": { "Sprint 1": "D", "Sprint 2": "R" }
}
]
}
Then run:
"$DELIVERABLE_ROOT/skills/deliverable/bin/excel-export" roadmap docs/requirements/roadmap-data.json
This writes docs/requirements/roadmap.xlsx.
"BRD complete. Ready for the technical spec? Say 'write technical requirements' to continue."
npx claudepluginhub canhta/deliverable --plugin deliverableProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.