What does this skill do, and when should you use it?
This Claude Code skill turns the AI into a naming strategist, running a seven-step process to name products, SaaS tools, brands, open source projects, and apps. It starts with a naming brief and explores metaphor territories instead of querying a thesaurus. Candidate generation, filtering, availability checking, and scoring all happen autonomously mid-process — the user only sees a vetted shortlist of 3-5 names with verified availability. The skill ships 14 on-demand reference files plus a bundled script that batch-checks domains, npm, PyPI, GitHub, Telegram, and more.
Establishes a naming brief by asking six questions (function, audience, feel, platform targets, etc.); loads only the needed files from 14 references (metaphor-mapping.md, principles.md, etc.) to explore metaphor territories and generate 30-50+ candidates; filters out AI slop like -ly/-ify suffixes via an anti-pattern checklist; runs the bundled scripts/check-availability.sh plus whois, curl, npm view, and WebSearch to actually verify availability on domains, npm, PyPI, GitHub, crates.io, RubyGems, WP plugin slugs, and Telegram; finally presents a weighted-scored shortlist with origin stories, availability status, and tagline suggestions.
How do you install this skill?
- Availability checks depend on whois and multiple overseas services (npm, PyPI, GitHub, t.me, etc.), which may be unreachable or time out from mainland-China networks, blocking or corrupting Step 5.
- The source of check-availability.sh is not included in this review's evidence; its behavior, exit codes, and false-positive modes cannot be statically verified — read it manually before use.
- Availability queries send candidate names to external services, potentially exposing not-yet-public naming intent; for confidential projects consider skipping or manually running checks.
- The skill is English-centric with no Chinese language file; applicability to Chinese-language brand naming is limited.
- No version numbers or changelog exist, so content changes between updates cannot be determined.
- Shell / CLI
- Network access
- Local filesystem
whoiscurlGitHub CLI (gh)npm CLI (for registry checks)
Clone the repo into a Claude Code skills directory. Project-level: mkdir -p .claude/skills && git clone https://github.com/glacierphonk/naming.git .claude/skills/naming. Personal (all projects): mkdir -p ~/.claude/skills && git clone https://github.com/glacierphonk/naming.git ~/.claude/skills/naming. Claude Code must already be installed.
git clone --depth 1 https://github.com/glacierphonk/naming.git ~/.claude/skills/namingGenerated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Type /naming in Claude Code and describe what you need a name for in one sentence, or just describe the challenge in conversation — Claude loads the relevant references automatically. Availability checking requires local commands (whois, curl, gh, npm) and network access; platforms the script doesn't cover (app stores, social handles) are checked via WebSearch. A quick session loads only 2-3 reference files.
What are this skill's strengths and limitations?
- Concrete methodology: 7-step process, weighted scoring rubric, anti-pattern checklist — not generic advice.
- Mandatory real availability checks using whois/curl/npm tools rather than guessing from memory.
- Well-designed context management: 15+ reference files load on demand; a simple task loads only 2-3 files.
- Loop-back mechanism: when candidates fail or score under 70, it returns to earlier steps instead of lowering the bar.
- MIT licensed and open source, welcoming language and industry guide contributions.
- Built for Claude Code — the /naming slash command and allowed-tools frontmatter need adaptation for other platforms.
- Availability checks depend on local commands (whois, curl, gh, npm) and network; without them, check quality degrades, and a whois failure falls back to merely testing whether a site is live.
- No test suite or sample end-to-end naming outputs are shown in the source, so real-world results are independently unverified.
- The core claim (real-word names ~68.8% recall vs ~38.1% for invented names) is cited without a source in the provided material.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| The Naming Strategist this page | 57 · Use with care | ★ 113 | 6mo ago | MIT |
| Brand Language Codifier | 63 · Recommended | ★ 3k | 11d ago | Apache-2.0 |
| Positioning Mapper | 57 · Use with care | ★ 3k | 11d ago | Apache-2.0 |
| Narrative Registry | 55 · Use with care | ★ 3k | 11d ago | Apache-2.0 |
| Claude Blog Brain | 53 · Use with care | ★ 2.4k | 11d ago | MIT |
How did FollowSkills review this skill?
allowed-tools are scoped to read-only operations (Read/Grep/Glob, whois, npm view, gh repo view, WebSearch/WebFetch); no writes, deletions, or credential access. Data flow is transparent: candidate names are sent via whois/curl/search to external services, clearly stated. Deductions: external network queries (which reveal the user's naming intent) require no explicit user confirmation; no explicit sensitive-data handling note; no rollback/recovery guidance; publisher unverified but not deducted per rules. ~8 points deducted.
Process is self-consistent: 7 steps, explicit loop-back conditions, a mandatory availability gate (Step 5) with fallbacks (curl when whois missing). Deductions: the bundled check-availability.sh source is not in evidence, so its behavior cannot be statically confirmed; no tests; thin edge/abnormal-input handling; whois/curl are network-sensitive and error-prone (availability.md itself admits false positives). Static cap of 10 applies; scored 9.
Trigger semantics are clear: description names target objects (products/SaaS/brands/open source/bots), Step 1 defines inputs, Step 7 defines output format, with boundaries implied. Deductions: skill is English-centric; language files cover only Polish and Portuguese — no Chinese support; core availability checks depend on whois and overseas services (npm, PyPI, GitHub, t.me, WordPress API) whose reachability from mainland China is questionable. ~6 points deducted.
Good layering: SKILL.md as entry point only, explicit context-budget management, on-demand reference loading table; README install docs; MIT license, branch protection, markdownlint and link-check CI, detailed CONTRIBUTING with maintainer response expectations (solo maintainer, 1-2 weeks). Deductions: no version number, no CHANGELOG, no known-limitations section. ~3 points deducted.
The core task (availability-verified finalists with origin stories) is well designed with genuine marginal value via AI-slop filtering and mandatory availability checks. Deductions: static review cannot verify actual output quality; final results still need user trademark/legal diligence; the checks themselves admit false positives requiring manual verification. Static cap of 7; scored 6.
Some auditable primary material: case studies require verifiable sources, CI includes lint and link-check workflows, contributing guide requires sourced facts. Deductions: no test suite, no execution evidence; the recall statistic (68.8% vs 38.1%) lacks a citable source; most reference files are not included in the evidence. Static cap of 5; scored 4.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →