How to use the Influencer Discovery via Apify agent skill

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md instruction file and optional scripts that an AI agent can load when a task matches its description. This skill tells the agent how to run an influencer data pull and how to handle the results responsibly.

Once installed, the agent can help you find and compare influencers on Instagram, Facebook, YouTube and TikTok. It picks a suitable Apify Actor (a scraper that runs as a paid cloud data-collection service), asks about output format and result count, then runs a bundled Node.js script that pulls public profile and post data into a chat summary, CSV or JSON file.

• Finds candidates by niche keyword, hashtag, or profile, depending on the platform Actor chosen.
• Reports metrics such as followers and engagement rate where the Actor returns them.
• Reminds the agent that follower and like counts are weak evidence of authenticity, not proof.

How do you install the Influencer Discovery skill?

Download the ZIP for this skill from this page and unzip it. Put the resulting folder in one of two places: your personal skills folder at ~/.claude/skills/, or a project-specific .claude/skills/ folder inside the project you are working in.

If you use the Claude apps instead of Claude Code, upload the skill under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/ (available across your projects).
• Project install: .claude/skills/ inside the project folder (available only there).
• Claude apps: Settings → Capabilities → Skills, then upload.

What do you need before running it?

Check these prerequisites before any paid run. The skill itself says to verify them first.

Your Apify token must be stored locally in a .env file, not pasted into the chat or into a command. The bundled runner reads it from there.

• Node.js 20.6 or newer (the commands use native --env-file support).
• An Apify account and API token, stored in a local .env file as APIFY_TOKEN=your_token.
• Apify connector or Apify Console access, so the agent can inspect the chosen Actor's input schema.
• Network access to the Apify service.
• Keep .env untracked and private, for example by keeping it out of version control.

How do you trigger the skill with a natural request?

You do not need special syntax. Describe the outcome you want in plain language and mention the platform and the kind of creator. The skill's description covers finding and evaluating influencers, assessing public authenticity signals, and tracking collaboration performance.

Expect the agent to ask follow-up questions before running anything: the output format (quick answer, CSV or JSON), the number of results, and the authorized public-data scope.

• Example: "Use the influencer discovery skill to find 20 fitness creators on TikTok for a spring campaign, and give me a CSV."
• Example: "Compare the public engagement of these three YouTube channels using the Apify skill. Quick answer is fine."
• Example: "Build a shortlist of Instagram creators in the sustainable fashion niche and tell me what public metrics you can and cannot verify."

How does a run work step by step?

The skill follows a fixed workflow. The agent first selects an Actor from a table of candidates for each platform and need, such as profile metrics, hashtag discovery, reel engagement or live streamers. It then inspects the Actor's current input schema through your connector or the Apify Console, because schemas can change. Treat the table as candidates, not as confirmed availability.

Next the agent asks your preferences and runs the script. The commands below show the three output modes. Replace ACTOR_ID and JSON_INPUT with the values the agent resolved, and use a new output filename each time. The placeholder for the skill folder is filled in by the agent.

• Quick answer: displays the top few results in chat and saves no file. It fetches at most five records.
• CSV: a bounded export with the union of returned fields.
• JSON: a bounded export in JSON format.
• Summary: the agent reports the number found, the file location, available metrics and suggested next steps.

What safety and cost limits should you know?

Apify runs are billed. The skill says to confirm the Actor, the public-data scope, the input schema, result limits and pricing before launch. Start with a small input limit supported by the Actor.

The --max-items option (default 1000) limits only the downloaded output. It does not cap the work the Actor does remotely or what you are billed. If a local timeout occurs, the remote run keeps going, so check the printed run ID in the Apify Console before retrying to avoid duplicate charges.

The skill is marked as running scripts and using the network. The helper sends a skill-identifying User-Agent to Apify. Do not collect private group data or unnecessary personal contact details, and minimise or redact personal fields before sharing results.

• Check the run's cost before scaling up.
• Keep output files at new paths; the skill requires new filenames.
• Label suspicious signals as uncertain. Do not accuse anyone of fraud based on scraped ratios.
• Do not use the data to contact creators without separately authorised outreach.

What tips help you get useful results?

Define the partnership brief and the comparison window before ranking anyone. The skill recommends deduplicating stable creator IDs and recording the date metrics were retrieved, the post sample size and the engagement denominator, so comparisons are fair.

Use the public numbers as a starting point for judgement, not as a verdict. Follower counts and likes cannot confirm real audience quality, demographics or return on a collaboration. For those, ask the creator for first-party evidence.

• Start small: a few records first, then widen once the output looks right.
• Record the retrieval date and post sample so you can explain your ranking later.
• Fix errors quickly: if APIFY_TOKEN is not found, create the .env file with APIFY_TOKEN=your_token; if an Actor is not found, check the Actor ID spelling.