By David Speakman ·
speakmanai-cc is the Claude Code skill layer for SPEAKMAN.AI: one skill per SDLC phase, each orchestrating a real multi-agent session and turning the result into something you actually keep. A document. A codebase. A mockup. A report. This update adds a phase on either end of the pipeline you already had, and gets the whole set onto Claude Desktop in a single upload.
/generate-project-brief runs entirely in Claude Code (no SPEAKMAN.AI session,
no credits), coaching a pitch, a Slack thread, a meeting note, or a half-formed idea into a
structured brief through guided discovery, in three modes: Create, Update, and Validate. Its coaching stance
and audit-trail structure are adapted, not copied, from the product-brief workflow in
BMAD-METHOD, an open source
AI-driven development framework with a contributor community in the tens of thousands, used
under BMAD-METHOD's MIT license.
The brief's own structure was built to be handed straight to Phase 1 or a sizing gate. No reformatting step, no glue code, no adapter, just a file path.
/generate-ux-design maps your architecture and requirements into a screen
inventory, checks that every requirement traces to a real screen, then proposes a design
direction and produces self-contained HTML mockups for the highest-priority screens. A
real thing to react to, not a paragraph describing one. The mockup step runs through your
own connected coding agent rather than a fixed background model, because design judgment is
exactly where that choice matters most.
The review gate is a real one. The skill publishes every mockup individually and waits for your actual answer: approve, or describe exactly what to change and watch it revise in place, not start over.
A raw idea, coached into a structured brief. Zero credits, no MCP session.
Business description in, Solution Architecture Document out.
Domain model, naming dictionary, use cases, API and DB schema.
Screen inventory, flow diagrams, a proposed direction, and clickable mockups.
A working, milestone gated, git committed codebase.
Terraform for GCP, AWS, and Azure, plus pre-pentest hardening.
A gated security engagement, run against the live staging build.
Run /sdlc-pipeline for a guided walkthrough that drives every phase in order and
carries state between them, checkpointed so a multi-day build survives a closed laptop.
Claude Desktop's Cowork mode runs skills in Anthropic's own sandboxed VM, which can't reach
a Claude Code install's asset directory. So every skill in the desktop bundle is fully
self-contained rather than depending on a separate setup step. Both new skills needed zero
adaptation to cross over: generate-project-brief only ever writes local files,
and generate-ux-design only ever calls MCP tools and Claude's own rendering
capability, so neither one touches anything sandbox-specific in the first place.
All 17 skills, one upload, via Cowork → Settings → Plugins.
No Cowork? Individual skills upload one at a time via Settings → Skills.
Every skill talks to the same SPEAKMAN.AI MCP server Claude Code uses: local or hosted, your choice.
Bumped from 14 skills to 17. See the plugin's own README for the update path.
Clone the repo and run the install script, or grab the Claude Desktop plugin. Either way, you're running against your own SPEAKMAN.AI deployment.
These skills work against any SPEAKMAN.AI deployment. The local engine is MIT licensed, no account required.
Written by David Speakman. Speakman Consulting designs and builds this kind of system for growing organizations: agent workflows with the governance that keeps humans in the loop.