RFP Workflow Skills

A no-code, zero-setup collection of AI assistant skills for analyzing RFPs and generating delivery artifacts. Use the skills inside GitHub Copilot, Cursor, and Claude Code with no hosted service or extra runtime.

Install with Skillfish

This repository is compatible with skillfish, the skill manager for AI coding agents.

Install the full pipeline skill:

skillfish add alxp1982/RFP_Workflow rfp-full-workflow

Install the umbrella workflow skill:

skillfish add alxp1982/RFP_Workflow rfp-workflow

Install all available skills from this repository:

skillfish add alxp1982/RFP_Workflow --all

Install with npx skills (skills.sh / Vercel skills CLI)

This repository matches the open agent skills layout (skills/<name>/SKILL.md), so it installs with the official skills CLI (docs).

List skills without installing:

npx skills add alxp1982/RFP_Workflow --list

Install the full pipeline skill (pick agents interactively, or pin e.g. Cursor + Claude Code):

npx skills add alxp1982/RFP_Workflow --skill rfp-full-workflow -a cursor -a claude-code -y

Install the umbrella workflow skill:

npx skills add alxp1982/RFP_Workflow --skill rfp-workflow -a cursor -y

Install every skill from this repo (non-interactive):

npx skills add alxp1982/RFP_Workflow --skill '*' -y

From a local clone (same as published GitHub layout):

npx skills add . --list
npx skills add . --skill rfp-full-workflow -y

The CLI writes into each agent’s configured skills directory (for example Cursor uses the paths described under Supported agents). This repo also keeps skills/ as the canonical copy and documents symlinks for other layouts in Multi-agent scaffold compatibility below.

Install with GitHub CLI skills

This repository is compatible with gh skill install via the standard skills/*/SKILL.md layout.

References:

• GitHub CLI gh skill docs: https://cli.github.com/manual/gh_skill
• GitHub CLI gh skill install docs: https://cli.github.com/manual/gh_skill_install
• Agent Skills specification (SKILL.md format): https://agentskills.io/specification

Install the full pipeline skill:

gh skill install alxp1982/RFP_Workflow rfp-full-workflow

Install the umbrella workflow skill:

gh skill install alxp1982/RFP_Workflow rfp-workflow

List available skills from this repository:

gh skill install alxp1982/RFP_Workflow

Multi-agent scaffold compatibility

skills/ is the canonical source of truth. For scaffold compatibility, these directories are symlinked to skills/:

• .agent/skills
• .claude/skills
• .cursor/skills
• .github/skills

Recommended mode: full pipeline (one shot)

Use the full pipeline skill (rfp-full-workflow) to execute the full chain with human-in-the-loop checkpoints. It starts with a Planning summary (goals, pipeline, checkpoints, risks) for alignment—similar to planning mode—then runs the stages.

Use #file:skills/rfp-full-workflow/SKILL.md

Input: [paste RFP text or #file:path]
PRD target: local-md | google-sheets | <mcp-target>
Stories target: local-md | github | jira | <mcp-target>

You do not need to invoke each skill manually.

What it does

Paste raw client requirements and run the skills in sequence to produce:

Artifact	Output
Clarification questions + assumptions	outputs/clarifications.md
Product Requirements Document	outputs/prd.md
Machine-readable PRD spec	outputs/prd.spec.yaml
Spec digest & changelog (agent context)	outputs/spec-digest.md, outputs/spec-changelog.md
Refined PRD (after answering questions)	outputs/prd.md (overwrite)
Architecture & technology stack options	outputs/architecture.md
Technical task decomposition	outputs/task-breakdown.md
Gherkin stories with acceptance criteria	outputs/stories.md
Machine-readable stories spec	outputs/stories.spec.yaml
New product repo kit (spec-driven baseline: spec/, AGENTS.md, CLAUDE.md, .cursor/, .claude/, .github/copilot-instructions.md)	outputs/repo-kit/ (copy tree to a fresh repo root)
Export to Google Sheets / GitHub / Jira	via rfp-sync-trackers skill

Bootstrapping a new product repository

After a full pipeline run, outputs/repo-kit/ contains a ready-to-copy tree for a new product git repository (see templates/repo-kit/README.md for the full layout: spec/, AGENTS.md, product-only Cursor/Claude rules and nested skills, Copilot instructions, optional docs/ guides, and so on). Those nested skills and docs/ are not part of this RFP Workflow repo’s skills/ pipeline—they ship only inside the kit for use after you copy repo-kit/ into the product repo root. Copy the contents of repo-kit/ into the root of a new empty git repository, commit, and start implementation with spec-driven agent defaults. The rfp-bootstrap-repo skill defines exact paths and placeholder rules.

Skills pipeline

RFP input (file or text)
       |
       v
rfp-normalize-rfp      -> normalize requirements
       |
       v
rfp-clarification-pass     -> assumptions + open questions (non-blocking)
       |
       v
rfp-draft-prd   -> full PRD markdown
       |
       v
rfp-refine-prd  -> refined PRD after answering questions  (optional)
       |
       v
rfp-architecture-stack -> architecture memo + stack comparison + selection
       |
       v
rfp-task-breakdown   -> Epic -> Feature -> Story -> Task hierarchy
       |
       v
rfp-user-stories     -> Gherkin-style stories + acceptance criteria
       |
       v
rfp-bootstrap-repo   -> portable `outputs/repo-kit/` for a new product git repo
       |
       v
rfp-sync-trackers      -> Google Sheets / GitHub Projects / Jira

How to use from your AI tool

GitHub Copilot (VS Code)

Reference a skill file in Copilot Chat and provide your input:

Use #file:skills/rfp-draft-prd/SKILL.md

Requirements:
[paste RFP text here]

Chain skills across a session using #file: for each stage. .github/copilot-instructions.md pre-loads the workflow context automatically.

Cursor

.cursor/rules/rfp-workflow.mdc loads automatically. Then:

Run the PRD draft skill on the following requirements: [paste text]

Or open any skill file and use Cmd+L to chat against it directly.

Claude Code

Reference the installable skill path directly:

Run #file:skills/rfp-draft-prd/SKILL.md on: [paste RFP text or reference a file]

Generic (any AI chat)

Copy a skills/*/SKILL.md file and paste it into the conversation, then append your requirements text.

Repository layout

skills/
  rfp-workflow/SKILL.md                Repo entry — points at full pipeline skill
  rfp-full-workflow/SKILL.md           One-shot pipeline with human checkpoints
  rfp-normalize-rfp/SKILL.md           Turn raw RFP into structured requirements
  rfp-clarification-pass/SKILL.md      Questions + assumptions (non-blocking)
  rfp-draft-prd/SKILL.md               Write PRD from requirements
  rfp-refine-prd/SKILL.md              Update PRD after clarification answers
  rfp-architecture-stack/SKILL.md      Architecture + stack options before breakdown
  rfp-task-breakdown/SKILL.md          Epic → Feature → Story → Task plan
  rfp-user-stories/SKILL.md            Gherkin-style acceptance stories
  rfp-bootstrap-repo/SKILL.md          Materialize `outputs/repo-kit/` for a new repo
  rfp-sync-trackers/SKILL.md           Push artifacts to Sheets / GitHub / Jira

templates/
  prd.md                     Canonical PRD template
  stories.md                 Canonical stories template
  clarifications.md          Clarifications report template
  task-breakdown.md          Task hierarchy template

.github/
  copilot-instructions.md    Copilot workspace instructions

.cursor/
  rules/rfp-workflow.mdc     Cursor rules file

docs/
  architecture.md            Pipeline design notes
  integration/
    google-sheets.md         Sheets export mapping
    github-projects.md       GitHub Projects export mapping
    jira.md                  Jira export mapping

examples/
  sample-rfp.md              Sample RFP input
  sample-outputs/            Expected output artifacts

Design principles

• No code to run. Every skill is a prompt + output template.
• Tool-agnostic. Works in any AI assistant.
• Non-blocking clarifications. Ambiguity becomes assumptions, never a hard stop.
• Fixed hierarchy. PRD → architecture & stack → Epic → Feature → Story → Task — every time.
• Traceable. Every artifact carries assumptions and open_questions.
• Flexible output targets. Default local markdown; export skills handle Sheets / GitHub / Jira.