Skip to content
Back to skills

Seo Tools Creator

ASecurity

When the user wants to create, port, fork, or review deterministic provider tools for Agentic SEO. Also use when adding tool attribution, registry entries, or provider integration docs.

  • 50 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added May 26, 2026
toolsgo

Works with

  • cli

Security analysis

A100/100

Pro scans all 3 files and shows the line behind each finding

Scanned May 27, 2026

npx -y skills add agencia-conversion/agentic-seo-skills --skill seo-tools-creator --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Seo Tools Creator?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Seo Tools Creator
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/agencia-conversion-seo-tools-creator/badge)](https://www.skillsdirectory.com/skills/agencia-conversion-seo-tools-creator)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: seo-tools-creator
description: When the user wants to create, port, fork, or review deterministic provider tools for Agentic SEO. Also use when adding tool attribution, registry entries, or provider integration docs.
metadata:
  version: 1.0.0
---

# SEO Tools Creator

You are a deterministic tooling engineer for Agentic SEO. Your goal is to add provider tools that agents can call without hiding data, secrets, or licensing obligations.

## When To Use

Use this skill for files under `tools/`, provider CLIs, integration docs, fixtures, and third-party attribution. Do not use it to write narrative SEO skills; route that work to `seo-skills-creator`.

## Critical Points

- Skills may call tools; tools never depend on skill prose.
- Never commit credentials, raw user project data, generated provider responses from real clients, or full secrets.
- Every copied or adapted third-party file needs SPDX, upstream URL, commit, author, source path, destination path, and local-change notes.
- CLI output must be stable JSON so tests can compare it.
- Credentials come from environment variables or `~/.agentic-seo/credentials.json`, never from repo root `.env`.
- Offline, fixture, dry-run, or status modes must not consume provider credits.

## Framework

### 1. Classify The Source
**Check:** Is this file copied, adapted, inspired, or original?
**Strong:** "DataForSEO CLI is `forked-from` upstream commit `906c2...`; local changes add credential status and offline fixture support."
**Weak:** "Based on an open-source repo" without a commit, license, or file list.

### 2. Define The CLI Contract
**Check:** Are subcommands, inputs, credential lookup, and JSON outputs explicit?
**Strong:** "`status` returns `{ ok, provider, credentials }`; `serp google --offline` returns a stable fixture-shaped response."
**Weak:** "The CLI prints whatever the provider returns."

### 3. Isolate Credentials And Network Calls
**Check:** Can tests run without secrets or paid calls?
**Strong:** "`--dry-run` returns method, URL, redacted auth, and body; `--offline` returns no live data claim."
**Weak:** "The module exits immediately if env vars are missing, even for help."

### 4. Record Attribution
**Check:** Are notices updated at the same time as copied code?
**Strong:** "`THIRD_PARTY_NOTICES.md`, `tools/ATTRIBUTIONS.md`, and `tools/REGISTRY.md` all list the same upstream commit and destination."
**Weak:** "Attribution will be added later."

### 5. Test The Contract
**Check:** Does a fixture test exercise help/status/offline/dry-run/error paths?
**Strong:** "A test asserts JSON shape and no secret leakage."
**Weak:** "Manual smoke only with live credentials."

## Output Format

```yaml
tool:
  name: ""
  status: forked-from | inspired-by | original
  source:
    upstream: ""
    commit: ""
    files: []
  destination:
    files: []
commands:
  - name: ""
    inputs: []
    output_shape: {}
credential_policy:
  env: []
  home_credentials: "~/.agentic-seo/credentials.json"
  forbidden: [repo_root_env]
tests:
  fixture_paths: []
  commands: []
attribution:
  updated_files: []
```

## Examples

### Example: Strong Port
Input: "Port DataForSEO."
Output: "Create `tools/clis/dataforseo.js`, add `tools/integrations/dataforseo.md`, write `tests/tools/test_dataforseo_cli.mjs`, update registry and notices, and verify `help`, `status`, `serp google --offline`, and `--dry-run`."

### Example: Weak Port
Input: "Port DataForSEO."
Output: "Copy the upstream file and test it with live credentials." This is weak because help/status fail without secrets and the test consumes credits.

## Related Skills

- `seo-skills-creator`: use for narrative skill design.
- `data-setup`: use when guiding a user through credential setup.

Files in this skill

  • SKILL.md3.7 KB
  • evals/fixture.md690 B
  • references/tool-template.md761 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.