Summary
Package Wiggum's core capabilities (project scanning, AI interview, spec generation) as installable skill files for coding agents like Claude Code, Gemini CLI, and Codex — turning every agent ecosystem into a distribution channel.
Problem / Context
Competitors require users to switch to them as the primary tool. Wiggum can take a fundamentally different approach: meet users where they already are. A developer using Claude Code who installs wiggum.skill.md gets /wiggum-scan, /wiggum-interview, /wiggum-spec inside their existing workflow — no context switch required.
This is both a strategic positioning play (Wiggum as the intelligence layer inside any agent) and a growth channel (users who discover Wiggum as a skill graduate to the full CLI). The skill.md can ship early as a parallel track since it packages existing capabilities.
Roadmap phase: Phase 3 — ENRICH (parallel track, can ship during Phase 1-2)
Proposed Solution
Phase 1: Claude Code skill.md
- Create
wiggum.skill.md that exposes key workflows as slash commands
/wiggum-scan — runs the project scanner and outputs a structured summary/wiggum-interview — runs the AI interview flow for a feature/wiggum-spec <name> — generates a spec from interview results- Skill file references
wiggum CLI (must be installed globally via npm i -g wiggum-cli)
Phase 2: Multi-agent skill distribution
- Adapt skill format for Gemini CLI, Codex, and other agents that support skill/instruction files
- Create an
npx wiggum install-skill command that detects the agent and installs the right format
- Document skill installation for each supported agent
Phase 3: Skill ecosystem
wiggum publish-skill — share custom skills- Registry of community-contributed skills
Acceptance Criteria
Summary
Package Wiggum's core capabilities (project scanning, AI interview, spec generation) as installable skill files for coding agents like Claude Code, Gemini CLI, and Codex — turning every agent ecosystem into a distribution channel.
Problem / Context
Competitors require users to switch to them as the primary tool. Wiggum can take a fundamentally different approach: meet users where they already are. A developer using Claude Code who installs
wiggum.skill.mdgets/wiggum-scan,/wiggum-interview,/wiggum-specinside their existing workflow — no context switch required.This is both a strategic positioning play (Wiggum as the intelligence layer inside any agent) and a growth channel (users who discover Wiggum as a skill graduate to the full CLI). The skill.md can ship early as a parallel track since it packages existing capabilities.
Roadmap phase: Phase 3 — ENRICH (parallel track, can ship during Phase 1-2)
Proposed Solution
Phase 1: Claude Code skill.md
wiggum.skill.mdthat exposes key workflows as slash commands/wiggum-scan— runs the project scanner and outputs a structured summary/wiggum-interview— runs the AI interview flow for a feature/wiggum-spec <name>— generates a spec from interview resultswiggumCLI (must be installed globally vianpm i -g wiggum-cli)Phase 2: Multi-agent skill distribution
npx wiggum install-skillcommand that detects the agent and installs the right formatPhase 3: Skill ecosystem
wiggum publish-skill— share custom skillsAcceptance Criteria
wiggum.skill.mdworks in Claude Code with/wiggum-scan,/wiggum-interview,/wiggum-spec~/.claude/skills/or equivalent)npx wiggum install-skillauto-detects agent and installs appropriate format