Back to list
clix-so

clix-skill-creator

by clix-so

An open collection of agent skills for Clix.

5🍴 0📅 Jan 16, 2026

SKILL.md


name: clix-skill-creator display-name: Skill Creator short-description: Generate new Clix agent skills description: Helps authors create new Clix agent skills by first researching the latest Clix SDK + docs via the Clix MCP Server, then generating a complete skill folder (SKILL.md, references, scripts, examples) aligned with the conventions in this repository. Use when the user asks to create/author a new Clix skill, extend the skills library, or when the user types clix-skill-creator. user-invocable: true

Clix Skill Creator

Use this skill to create a new Clix skill that matches a user's need, while avoiding duplicated scope and preventing hallucinated API usage by relying on the Clix MCP server as the source of truth.

This is a “meta-skill”: it does not integrate an SDK directly; it guides you to produce another skill folder (docs + deterministic scripts) that does.

When to use this skill

  • User says: “Create a new skill for Clix that helps with X”
  • Internal request: “We need a new skill for feature X in Clix”
  • User provides a workflow that is not well-covered by existing skills
  • User types: clix-skill-creator

Non-goals (important)

  • Do not create a new skill if the request can be satisfied by combining existing skills (integration + event tracking + personalization + user management + API-triggered campaigns).
  • Do not invent SDK methods, endpoints, or constraints. If you can’t confirm via MCP, treat it as unknown and ask the user or add a TODO note.
  • Do not embed secrets, project IDs, API keys, or customer data in the skill.

MCP-first (mandatory source of truth)

Before writing any new “Clix API behavior” or “SDK signature” into a newly generated skill:

  • MUST use clix-mcp-server:search_docs to confirm conceptual behavior, limits, and console semantics.
  • MUST use clix-mcp-server:search_sdk to confirm exact method signatures for each platform (iOS/Android/Flutter/React Native) that the new skill will mention.

If clix-mcp-server tools are not available:

  • Ask the user whether to install the MCP server (prefer using the existing installer in skills/integration/scripts/install-mcp.sh).
  • If the user declines, proceed with an explicit “static fallback may be outdated” warning and minimize hard claims.

Workflow (copy + check off)

Clix skill creation progress:
- [ ] 1) Intake: user need → crisp problem statement + acceptance criteria
- [ ] 2) Scope check: confirm this isn’t already covered by existing skills
- [ ] 3) MCP research: gather evidence (docs + SDK signatures) for the target scope
- [ ] 4) Draft a Skill Brief (name, trigger phrases, inputs/outputs, guardrails)
- [ ] 5) Generate scaffold (SKILL.md + references/ + scripts/ + examples/)
- [ ] 6) Validate scaffold (structure + frontmatter + MCP-first section)
- [ ] 7) Wire-in (README + llms index + tests if needed)

1) Intake: minimum questions

Ask only what’s needed to define a stable skill boundary:

  • Goal: what outcome should the skill reliably produce?
  • Audience: app devs, backend devs, marketers/ops, or mixed?
  • Platform (if SDK-related): iOS / Android / Flutter / React Native
  • Clix capability: push / in-app / email / audiences / journeys / etc.
  • Constraints: PII policy, compliance, performance limits, rate limits
  • Success criteria: what does “done” look like?

2) Reference official skills (match repo style)

Before creating a new skill, use the existing official Clix skills in this repo as your style guide so the generated skill matches the standards here (format, tone, workflow, validators), while still being tailored to the user’s need.

  • Study how they are written:

    • YAML frontmatter conventions
    • MCP-first “source of truth” behavior
    • Progressive disclosure (references/, scripts/, examples/)
    • Plan artifacts + deterministic validators (bash scripts)
  • Then generate the new skill based on the user’s need using the same patterns. If the need is clearly a tiny addition to an existing skill, prefer adding a references/ doc or examples/ file there instead of creating a new skill.

  • clix-integration

  • clix-event-tracking

  • clix-user-management

  • clix-personalization

  • clix-api-triggered-campaigns

3) MCP research: build an “Evidence Pack”

Create a short evidence pack that you will cite while writing the new skill:

  • Docs evidence (from clix-mcp-server:search_docs)
    • behavior guarantees
    • constraints/limits
    • console terminology
  • SDK evidence (from clix-mcp-server:search_sdk)
    • exact signatures per platform
    • initialization / required params
    • error behavior (throws? returns promise?)

Store the evidence in your notes (or as a references/ markdown file in the new skill), with the search queries you used so future maintainers can refresh it.

4) Draft a Skill Brief (output format)

Produce this brief for approval before generating files:

skill:
  folder_name: "<kebab-case>" # e.g. "push-troubleshooting"
  name: "clix-<kebab-case>" # e.g. "clix-push-troubleshooting"
  display_name: "<Title Case>"
  short_description: "<short>"
  description: "<2-3 lines>"
  user_invocable: true

triggers:
  - "phrases the user might say"

inputs:
  - "minimal inputs the skill will ask for"

outputs:
  - "artifacts the skill produces (plan JSON, code changes, checklists)"

guardrails:
  - "MCP-first requirements"
  - "security / PII constraints"

files_to_generate:
  skill_md: true
  references:
    - "<doc>.md"
  scripts:
    - "<script>.sh"
  examples:
    - "<optional>"

5) Generate scaffold (repo conventions)

In this repository, a “complete” skill folder should include:

  • SKILL.md (with YAML frontmatter; MCP-first; workflow; progressive disclosure)
  • LICENSE.txt
  • references/ (markdown docs; not empty)
  • scripts/ (deterministic bash scripts; not empty)
  • examples/ (optional, but recommended when copy/paste code is useful)

6) Validate scaffold (fast feedback loop)

After generating a new skill folder, validate it:

bash <skill-dir>/scripts/validate-same-scope.sh path/to/installed/skill-creator path/to/new-skill-folder
bash <skill-dir>/scripts/validate-skill-location.sh path/to/new-skill-folder --mode repo
bash <skill-dir>/scripts/validate-skill-scaffold.sh path/to/new-skill-folder

This should check:

  • the new skill is installed at the same scope as skill-creator (project-level vs user-level), i.e. the new skill folder lives next to the installed skill-creator under the same .../skills/ directory
  • the skill folder is in the correct skills/<name>/ location (for this repo)
  • required files exist
  • references/ and scripts/ are present and non-empty
  • SKILL.md has valid frontmatter keys
  • the skill includes an MCP-first section referencing clix-mcp-server

Progressive Disclosure

  • Level 1: This SKILL.md (always loaded)
  • Level 2: references/ (load when authoring the new skill)
  • Level 3: examples/ (load when copy/pasting scaffolds)
  • Level 4: scripts/ (execute directly; do not load into context)

References

  • references/skill-template.md
  • references/mcp-research-playbook.md

Score

Total Score

55/100

Based on repository quality metrics

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

1ヶ月以内に更新

+10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

+5

Reviews

💬

Reviews coming soon