Agent Skills by jal-co for deciding what to build — and how. Badges by shieldcn.dev. Per-skill install counts populate after skills.sh indexes each skill (first npx skills add).

The hard part of building isn't building anymore — execution is mostly handled. The bottleneck moved upstream, to the decisions: which idea is worth chasing, and what to build it with. This is a small collection of skills for exactly those calls.

A realistic co-founder critic for software ideas. You bring an idea, it tells you whether to ship it, shape it, or skip it.

I have a lot of ideas. What I don't have is the headspace to figure out which ones are worth chasing.

The hard part of building isn't building anymore. Models got good enough that execution is mostly handled, so the bottleneck moved. Now it's finding an idea that hasn't already shipped, in a niche big enough to matter, that someone will actually pay for.

This is the gut-check I wanted. You bring an idea, it does the research instead of guessing, and it tells you honestly where the competition is too big, the niche too small, or there's no path to revenue. When tokens are the only currency, your ideas matter.

The two pair up: ship-or-skip decides whether to build, stack-up decides what to build it with — and both append to the same docs/<idea-slug>.md.

npx skills add jal-co/skills --skill ship-or-skip

Works in any Agent Skills–compatible client (Claude Code, Cursor, Codex, Gemini CLI, opencode, pi, and more). It benefits most from an agent with live web search for current competitor, pricing, market, and domain data.

Manual install (Claude Code):

cp -r skills/ship-or-skip ~/.claude/skills/

Describe an idea, or name a module:

• "Ship or skip: an AI tool that writes podcast show notes." → full triage + scorecard + verdict
• "Who are the competitors for this?" → competitive landscape
• "Critique this idea hard." → deep teardown
• "Sharpen the positioning." → wedge & niche
• "Name this product and check domains." → naming + availability

Every module appends to one report per idea at docs/<idea-slug>.md, so your evaluations accumulate into a paper trail (shared with stack-up).

1. Problem severity
2. Market size / niche viability
3. Competition intensity (too crowded and too empty are both bad)
4. Differentiation / moat
5. Willingness-to-pay / monetization
6. Distribution & GTM reachability
7. Founder-fit / unfair advantage
8. Name & domain availability (lighter weight)

The score informs a holistic verdict — it doesn't mechanically decide it.

Once an idea is worth building, decide what to build it with:

• "What stack should I use for this?" → full stack recommendation across every layer
• "Postgres vs Mongo for this?" → focused head-to-head for your context
• "Audit my repo's stack." → flags legacy / unmaintained choices + modern alternatives
• "Give me the setup plan." → ordered scaffold commands and first-deploy steps

stack-up appends to the same docs/<idea-slug>.md, so one idea ends up with its verdict and its stack in a single document.

skills/                           # repo root (jal-co/skills)
├── README.md                     # this file
├── AGENTS.md                     # guide for authoring skills here
├── skills.sh.json                # skills.sh grouping metadata
├── LICENSE
└── skills/
    ├── ship-or-skip/
    │   ├── SKILL.md              # metadata + instructions
    │   ├── references/
    │   │   ├── scorecard.md      # the 8-axis rubric
    │   │   ├── research.md       # search-first research playbook
    │   │   └── naming.md         # naming + domain availability
    │   ├── assets/
    │   │   └── report-template.md
    │   └── scripts/
    │       └── check-domain.sh   # best-effort whois/dig domain check
    └── stack-up/
        ├── SKILL.md              # metadata + instructions
        ├── references/
        │   ├── stack-layers.md   # the layers to cover + right-sizing
        │   ├── requirements.md   # what to ask before recommending
        │   └── research.md       # search-first ecosystem research
        ├── assets/
        │   └── stack-template.md
        └── scripts/
            └── scan-stack.sh     # best-effort manifest/dependency scan

MIT — see LICENSE.