// Agent Skills

SEO Agent Skills for research and content decisions

A Skill tells an AI agent how to scope a research task, reuse prior work, handle evidence and report limitations. The instructions are open source, readable before installation and separate from the data service.

01 / Skill guide

$research-seo-signals

Research SEO signals

Collect the smallest sufficient SEO data scope for a keyword, domain, market and language, then report observations, evidence and limitations without making the final priority decision.

When to use it

Use it when you need underlying SEO observations for one defined keyword, domain, market and language before making a prioritization decision.

What the workflow enforces

  • Reuse a matching request_id and a stable idempotency key instead of creating duplicate research jobs.
  • Request only the evidence families the task needs, then read observations, evidence and limitations together.
  • Keep sourced facts separate from inference and stop before making the final SEO decision.

Natural-language example

Use $research-seo-signals to research “AI SEO tools” for example.com in the US English market, using keyword overview and current SERP evidence.

Where it stops

This Skill retrieves and explains SEO evidence. It does not guarantee ranking, turn demand into revenue, or decide whether the user should build the product.

02 / Skill guide

$research-social-signals

Research social signals

Turn a broad listening goal into focused platform-native retrieval, preserve source context and native fields, and disclose coverage limitations without turning retrieval into a recommendation.

When to use it

Use it when you need traceable public posts, notes, profiles, account content or trends from supported social platforms for downstream analysis.

What the workflow enforces

  • Translate a broad goal into a small focused query set and validate unfamiliar platform identifiers before retrieval.
  • Preserve source URLs, timestamps, native metrics, query context and pagination state; keep missing values unknown.
  • Return the retrieved data and its coverage limits without adding sentiment, strategy or business recommendations.

Natural-language example

Use $research-social-signals to find recent X conversations about PDF translation tools that preserve layout, with source links and coverage limitations.

Where it stops

Retrieved public content is partial platform evidence, not a representative market survey. The Skill does not score accounts or recommend what to publish.

03 / Skill guide

$decide-content-opportunities

Decide content opportunities

Use completed evidence to return one conditional stance with qualitative confidence, counter-evidence, risks, alternatives and a smallest useful next test.

When to use it

Use it after the decision, audience and constraints are clear and the available evidence can support a conditional keyword or content recommendation.

What the workflow enforces

  • Reuse completed evidence and compare materially different topics, audiences, angles, formats or channels.
  • Choose one primary action and show its rationale, counter-evidence, conditions, risks and alternatives.
  • Express confidence qualitatively and end with the smallest useful next test and a stop condition.

Natural-language example

Use $decide-content-opportunities to compare the completed SEO evidence and recommend one content direction, with confidence, counter-evidence, alternatives and a next test.

Where it stops

This Skill supports a reversible decision. It does not write or publish finished content, fabricate product experience, or replace the user's final business judgment.

// Open source

Inspect the complete Skill instructions

Review every SKILL.md file, its scope rules and version history in the public repository before you install it.

View all Skill source files on GitHub

// Install

Download and install SignalDig Agent Skills

Four steps: download, extract, copy, verify. No Git or extra tools required.

01Download the latest release

Click the button below to open GitHub Releases, then download the latest Source code (ZIP) archive to your computer.

Download the latest release
02Extract and enter the folder

Extract the downloaded ZIP, then enter the extracted folder in your terminal (all following commands run from here):

unzip seo-signal-skills-*.zip && cd seo-signal-skills
03Copy the Skill directories

Choose your AI client, then copy its command into the terminal from the previous step. The command creates the target directory and copies the three Skills into it.

Codex
mkdir -p ~/.codex/skills
cp -R skills/research-seo-signals ~/.codex/skills/
cp -R skills/research-social-signals ~/.codex/skills/
cp -R skills/decide-content-opportunities ~/.codex/skills/

The commands above work on macOS and Linux. On Windows: open the extracted folder in File Explorer and manually copy the three subfolders inside skills/ into your client's directory.

04Restart the client and verify

Restart the client or open a new session. You can also run the command below to confirm the three Skill directories are in place:

ls ~/.codex/skills

Install one Skill or all three as needed. After installation, invoke them in the client with the $skill name.

Install from source
git clone https://github.com/daily-growth-signals/seo-signal-skills.git
cd seo-signal-skills

A Git clone is useful when you want to inspect changes or update with git pull. It is not required for the release ZIP path.

// Use

Ask for a named workflow with a concrete task

Use $research-seo-signals to research “AI SEO tools” for example.com in the US English market, using keyword overview and current SERP evidence.
A clear keyword, domain, market and language help the workflow choose the smallest useful request. Generic requests may require a scope question before data is collected.

// Skills and MCP

Two layers, two different jobs

Keeping the layers separate makes the workflow inspectable while authenticated data capabilities can be governed and updated independently.

Provides

Task instructions, scope rules, evidence handling and output structure

Authenticated tools that retrieve current data or run a SignalDig report

Lives in

A local SKILL.md directory you can read and version

A remote HTTP endpoint configured in your AI client

Does not

Contain live market data or API credentials

Decide what your agent should ask or replace your business judgment

Need current SignalDig data in the workflow?

Continue to the MCP setup guide for endpoints, API-key authentication and current capability limits.