claude

product-workflows

Seventeen slash commands for defining a product — its requirements, architecture and specification: a six-command BRD-to-PRD route (/brd-intake → /brd-split → /prd-ground → /brd-split → /brd-interview → /brd-package → /brd-reconcile) that grounds a customer's requirements document, settles it with them and seeds the PRD, ARD and specification; the idea→PRD→ARD→specification ladder (/idea → /create-prd → /update-prd → /create-ard → /specify), /update-spec revising a merged specification, and /prd-ground optionally grounding the PRD; /epics, deriving Epics; /harvest-decisions and /promote-decisions, keeping merged ARD decisions as team records and promoting them to organisation ADRs; and /prd-proposal and /brd-proposal, pricing graded requirements in human hours, never money. Fourteen agents carry the grounding, review and Epic writing; twelve reference files define the shared formats.

2 on GitHubby Ivan Gudakv3.79.4MITupdated todaydevelopment
/

Plugin product-workflows from marketplace shipwright in ihudak/ai-workflows. Commands read from the repo's manifest — today.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Slash commands (17)

  • /product-workflows:brd-intake

    Starts the BRD-to-PRD route: takes a customer's business requirements document (markdown only), walks its links read-only, asking before it captures anything outside the document's folder, copies the document and what it links into the specs repo byte for byte, transcribes its images, and extracts a

  • /product-workflows:brd-interview

    The BRD-to-PRD route's decision step, once per slice after /prd-ground and /brd-split: generates the round's questions and tags each [G] (answered from the grounding, never asked), [V] (asked of you one at a time, with argumentation) or [C] (held for the customer), then writes decisions.md ([VD#n],

  • /product-workflows:brd-package

    The BRD-to-PRD route's customer-packaging step, once per slice after its interview round settles: runs an adversarial self-review and refuses to build while any [SR#n] finding is undisposed, then renders a self-contained customer prompt with the review schema inlined, a short delivery note and a pla

  • /product-workflows:brd-proposal

    Writes the programme effort proposal for a BRD container by rolling up its slices' own proposals, first saying whether the umbrella on disk is still current: walks each slice to a recommendation (price it, exclude and disclose it, re-run a stale one, include a current one), adds cross-slice effort n

  • /product-workflows:brd-reconcile

    The BRD-to-PRD route's last step: commits the customer's returned review into the BRD folder, reads it against the review schema or as free text (each free-text reading confirmed with you against its quotation), freezes the answers as [CD#n], closes the customer questions, applies the required corre

  • /product-workflows:brd-split

    The BRD-to-PRD route's allocation step. On a BRD root it carves slices from an instruction you type after the key, keys each as a PRD- folder inside the BRD, and walks every unallocated coverage-ledger row to a slice, a deferral, a rejection or superseded; on a slice it allocates that slice's own ro

  • /product-workflows:create-ard

    Authors an architecture requirements and decision document (ard.md) for a PRD or an Epic (/create-ard <KEY>; an Epic-level ARD inherits the PRD-level one), grounded in the mounted code repositories and, optionally, the architecture repository, gated by the Opus ard-reviewer. Optional; product-archit

  • /product-workflows:create-prd

    Authors a Product Requirements Document (prd.md) from a refined idea.md and a key, through a relentless grill against workflows-core:prd-format (--lean, --hybrid or --full profile), gated by the Opus prd-reviewer; product altitude, no code scan. On the BRD route it seeds from a decided slice's decis

  • /product-workflows:epics

    Drafts the child Epics of a PRD (a PRD- folder holding prd.md) or re-refines one Epic (an EPIC- folder holding epic.md under a PRD), from the PRD, the existing Epics and an optional code scan, each Epic with testable acceptance criteria and, where the run has a component set, one target component. R

  • /product-workflows:harvest-decisions

    Harvest the decisions of every ARD on the specs repository's default branch into the team architecture knowledge base at $SPECS_PATH/architecture/ — one record per [AD#N], reconciled on every run (superseded, withdrawn, applied in, deviated in), never deleted, handed off by pull request. Determinist

  • /product-workflows:idea

    Refines one source — an inline prompt or a markdown file (a note, a saved community post, product feedback, an existing PRD; its links and images walked, asking before it goes deep) — into a lean one-page idea.md through a bounded one-question-at-a-time grill (--deep for a relentless one). Copies th

  • /product-workflows:prd-ground

    Grounds a BRD slice's [BR#n] rows, or an idea-route PRD's own requirements, against code and design at pinned commits; every finding is re-derived blind by grounding-verifier. Type it, or let /brd-split's Ground now run it — any other model-started run stops. Next: /brd-split; on the idea route, /cr

  • /product-workflows:prd-proposal

    Writes a customer-facing effort proposal for one PRD- folder: grades it against four readiness tiers, derives work packages by delivery seam with hours by package and role and a bottom-up range, cites a verified finding, a frozen decision or a confirmed defect for every cost driver, and adds a defec

  • /product-workflows:promote-decisions

    Promote team architecture decisions to organisation ADRs. A product architect runs it inside the architecture repository. It ranks the live records of $SPECS_PATH/architecture/ — decisions several teams made, cited across PRDs, rarely departed from — and the accepted ADRs teams keep departing from.

  • /product-workflows:specify

    Authors a specification.md for a PRD or an Epic through a relentless one-question-at-a-time grill, lightly grounded in code, gated by the Opus spec-reviewer, and lands it on the specs repo's default branch by branch and pull request, ready for /dev-workflows:design. On the BRD route it seeds from th

  • /product-workflows:update-prd

    Updates an existing PRD through a relentless grill against workflows-core:prd-format, diffing against the current copy and drawing on the ARD, specification, transcripts and grounding findings, gated by the Opus prd-reviewer. Keeps every requirement id (a changed requirement gets a new one; a droppe

  • /product-workflows:update-spec

    Revises a merged specification.md in place, for a PRD or an Epic, grilling only what changed: what moved upstream since the inputs it was built from, what the operator says is wrong, and the notes /design, /implement and /conformance left. Keeps every ID (a change of meaning takes a new ID and super

Subagents (14)

  • ard-reviewer

    Reviews an Architecture Requirements/Decision Document (ARD) authored by /create-ard for grounding integrity (every as-is claim cites a real file:line), AD#N well-formedness (Binds/Prevents/testable Rule/Alternatives), supersession integrity (a changed decision superseded, never rewritten in place),

  • brd-package-reviewer

    Adversarially reviews a BRD package before it goes to the customer — attacks the position rather than summarising it, and returns [SR#n] findings each requiring a disposition. Read-only. Uses Claude Opus.

  • brd-reader

    Extracts a [BR#n] requirement inventory from a customer BRD — the document, the markdown /brd-intake copied, and its image transcriptions — one row per obligation, with a source anchor and unconfirmed defect candidates; raises an ambiguity on an obligation only an image states, and a conflict where

  • code-grounder

    Grounds specific requirement claims (a BRD's [BR#n] rows, or a PRD's [AC#n]/[FR#n]/[US#n] rows) against a single code repository at a pinned commit — one [CG#n] finding per claim, with file:line evidence and a verdict from the closed set. Answers "is this claim true of this commit?", not "what capab

  • customer-review-reader

    Reads a returned customer review of a BRD package in two modes — schema mode parses a file written against the customer review schema; free-text mode drafts that same schema from prose and emits every inferred decision as an unconfirmed candidate. Never promotes an inference to a customer decision.

  • design-grounder

    Reconciles a requirement inventory (a BRD's [BR#n] rows, or a PRD's [AC#n]/[FR#n]/[US#n] rows) against an exported design frame set — one [DG#n] finding per divergence, in four classes: a frame shows a field no requirement asks for; a requirement asks for a field no frame shows; a frame contradicts

  • epic-reviewer

    Reviews Epic drafts written by /epics for goal clarity, acceptance-criteria testability, scope boundaries, cross-Epic dependencies, and non-duplication with existing Epics under the parent PRD. Returns PASS / PASS WITH RECOMMENDATIONS / BLOCK. Uses Claude Opus. Product documentation is reviewed by d

  • epic-writer

    Writes child Epic-definition files for /epics from a structured handoff file — one file per Epic, following the Epic template, traceable to the folder read handoff and code-scanner evidence. Write-only — writes into the PRD folder, never commits (still true — it runs no git at all). Returns the list

  • figure-reader

    Transcribes the images a customer's BRD or an idea source links — every legible label, header, field, legend and annotation verbatim, what each annotation points at, and a diagram's flow — plus one sentence on what each image depicts. Reads only the images it is handed and never the document that li

  • grounding-verifier

    Independently re-derives grounding findings — [CG#n] from one pinned repository, [DG#n] from one exported frame set — in two dispatches: derive mode, blind to each finding's answer (it refuses a dispatch that carries one), then compare mode, returning agree, extend, contradict or unprovable against

  • idea-reader

    Turns one idea source (an inline prompt, or a markdown file such as a note, a saved community post, product feedback or an existing PRD) into a structured source digest for /idea. For markdown it reads exactly the pages and image transcriptions its caller hands over, maps them onto the digest's link

  • prd-reviewer

    Reviews a Product Requirements Document (prd.md) authored by /create-prd or /update-prd for goal crispness, user-story/acceptance-criteria testability, scope concreteness, internal consistency (no self-contradiction), measurable metrics, product-level purity (no implementation detail), downstream-co

  • proposal-reviewer

    Adversarially reviews an effort proposal and its rationale brief before either reaches a customer — re-adds the arithmetic, resolves every cost driver's citation against the record that owns its class, and checks the tier claimed against the evidence on disk. Read-only. Uses Claude Opus.

  • spec-reviewer

    Reviews a product specification.md authored by /specify or revised by /update-spec for per-stage quality (problem/scope/user-stories/acceptance-criteria/test-cases), cross-stage consistency, coverage, and identifier integrity. Read-only; returns findings + a PASS / PASS WITH RECOMMENDATIONS / BLOCK

Hooks (1)

  • UserPromptSubmit

    bash "${CLAUDE_PLUGIN_ROOT}/hooks/preload-context.sh"

View source on GitHub

More from ihudak/ai-workflows

All 8 plugins →

Shared foundation for the workflow plugin family: the addressing grammar and specs-repo git entry points, phase handoff, model routing, escalation and finding-triage rules, documentation and architecture grounding, cost/feedback/follow-up emission, and the reference loader skill that lets a dependent plugin read them. Also carries the family-meta utility commands — /feedback, /prompt, /prompt-brainstorm, /prompt-grill-me, /statusline — /frames for indexing design frame sets, and /diagnose-session for reading a session's transcripts to report, with evidence, why a run went wrong.

3 skills · 7 commands · 11 agents · hooks

2

Six slash commands for the implementation half of the product lifecycle: `/design` turns a specification into a design, with Opus review and an optional three-take fan-out on a contested interface. `/implement` builds it, gated by Opus plan critique and code review before tests. `/ready` verifies an ARD/spec/design chain's readiness, read-only, no tracker read. `/conformance` traces every requirement against an Epic's branch, or each repository's default branch, and records how each discrepancy is settled. `/upgrade` plans and executes dependency upgrades; `/vuln` researches and fixes a CVE. Fourteen subagents, one hook, nineteen reference files, no bundled skill. Every code-changing command commits before the run ends; push and PR sit behind one consent choice. Depends on the companion `workflows-core` plugin. Neither `product-workflows` nor `docs-workflows` is a dependency.

6 commands · 14 agents · hooks

2

Commands for a product's documentation and release notes. /docs-init scaffolds a documentation repository from nothing: a Material for MkDocs skeleton, public and internal builds over one content root, Vale, CI and a docs-profile.yml. /docs-brand extracts a logo and colour pair from the product's code and applies them, WCAG contrast-checked. /docs-serve runs the site's dev server. /docs-audit turns the product's own code and specs tree into a coverage grid and a prioritised backlog of the pages it is missing, behind an Opus review gate. /document edits prose directly, or keyed to a PRD folder writes product docs from its implementation diffs behind an Opus review gate. /docs-profile refreshes a docs repository's machine-readable profile; /release-notes drafts a destination-shaped entry from the same folder. Depends on workflows-core and prose-style.

1 skill · 7 commands · 11 agents · hooks

2

Pluggable prose style checker and writing aid for documentation, planning documents, and release notes. Ships a vendor-neutral baseline — terminology, word list, voice and tone, grammar, formatting, UI interactions, inclusive language, and prose accessibility — distilled from the Microsoft, Google, and Apple style guides, Chicago, the Conscious Style Guide, and WCAG. Point it at your own organization's style guide and the overlay augments or overrides that baseline; every resolution miss falls back silently. Provides /prose-review-pr, /prose-review-docs (with --fix), /prose-style-refresh, a prose-style-checker agent, a prose-fixer agent, and a prose-style-rules skill.

1 skill · 3 commands · 2 agents

2

Reviews an OpenAPI specification against bundled REST API and IAM permission-naming guidance, and reviews application code or UI against bundled public design-system and accessibility standards. Ships the guidance it reviews against — Google AIP, Zalando, Microsoft REST, OpenAPI 3.1 and ~20 RFCs for APIs; Apple HIG, Material Design 3, Microsoft Fluent 2, W3C WCAG 2.2, ARIA APG and NN/g for interfaces — plus a Spectral ruleset that runs against a spec directly. Standalone: it consumes no workflow artefact and produces none.

2 commands · 2 agents

2

acli

claude

Atlassian CLI (acli) reference skill for Jira and Confluence: JQL search, work items, comments, links, attachments, watchers, projects, sprints, boards, filters, and Confluence spaces, pages and blog posts. Adds container-friendly authentication (token on stdin; --web OAuth cannot complete without a browser) and a write policy in which --yes carries the user's approval, given in the conversation first, and a blast-radius count precedes any --jql/--filter mutation. Requires the acli binary on PATH.

1 skill

2

Similar plugins

superpowers

claudecodex

Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques

15 skills · hooks

obra/superpowers 316,836

context7

claudecodex

Up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your coding agent's context.

1 MCP server

upstash/context7 64,709

Build powerful, stateful CLI interfaces for any GUI application using the cli-anything harness methodology.

5 commands

HKUDS/CLI-Anything 58,665

Automates browser interactions for web testing, form filling, screenshots, and data extraction

12 skills

vercel-labs/agent-browser 50,373

Use Alibaba Open Code Review from local Codex.

2 skills

alibaba/open-code-review 50,068

slidev

claude

Claude Code skill for creating and presenting Slidev slidedecks

1 skill

slidevjs/slidev 49,452

Maintain product-workflows?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins8 plugins