How to use Claude
skills for talent acquisition.

A Claude skill is a folder with a SKILL.md file that teaches Claude a procedure for one specific job-to-be-done. The first lines of the file — the frontmatter — describe what the skill does and when to use it. Claude reads that description, decides when it’s relevant, and pulls in the rest of the file on demand.

The shift isn’t just better prompts — it’s how the model finds them. You describe the work in plain English, Claude recognises the pattern, and the right skill activates without you needing to ask. The output is tighter because Claude is executing a deliberate procedure — voice, structure, anti-patterns, and quality bars — instead of guessing.

Open the skills index and pick the moment in your funnel that’s breaking. Each skill carries an editorial tag for the stage it belongs to.

• Briefing. Role intake, ICP, and EVP — the upstream work every other skill reads from.
• Sourcing. Talent market map, then ranked boolean and filter strings.
• Outreach. First-touch and a follow-up sequence.
• Interviewing. Recruiter screen script and a competency scorecard.
• Closing & declining. Offer-stage playbook and stage-by-stage rejection copy.

Click any card and download the SKILL.md. You don’t need to start with all eleven — pick one stage you’d like to tighten this week.

Skills are an Anthropic platform feature. Enable them once and they’re available to every conversation in your workspace.

• Pro & personal. Open Settings → Capabilities in claude.ai and turn on Skills.
• Team & Enterprise. Your workspace admin enables Skills org-wide first — then individuals can turn them on.
• Claude Code. Skills are on by default. Drop the SKILL.md into ~/.claude/skills/<skill-name>/ and Claude Code picks it up without restart.

In claude.ai, open the Skills panel and add the SKILL.md you just downloaded. If a skill ships with supporting files — templates, examples, reference notes — upload the whole folder; Claude treats them as the skill’s working set and only loads them when needed.

Once uploaded, the skill is available to every conversation in your workspace. You don’t re-attach the file each time, and you don’t need to tell Claude it’s there.

This is the part that surprises people: you don’t tell Claude which skill to use. You describe the work, and Claude picks the right skill based on the description in the frontmatter.

• “Help me build a brief for our first VP Sales hire.” → role-intake-brief activates.
• “Build me a sourcing search for a Series B Head of Product in Berlin.” → sourcing-search-builder activates.
• “Draft a first-touch InMail for Maya — profile below.” → outreach-first-touch activates.

You can force a specific skill if you want — just name it (“use the role-intake-brief skill”) — but most of the time, Claude’s pick is right.

Skills are deliberately generic until you ground them. The output is only as good as the inputs you provide. For a typical talent acquisition skill, that means:

• The role title, company stage, and 90-day outcomes you want.
• Your existing role brief or ICP, if you have one.
• Comp band you can pay (cash + equity).
• Anti-patterns from past hires that didn’t work.

The skill will usually ask follow-up questions if a critical input is missing — that’s by design. Answer them honestly, even when the answer is “I don’t know yet.”

The first output is a working draft, not a deliverable. Stay in the same thread and tighten it:

• “Tighten the must-haves to four — drop anything we can train.”
• “Rewrite the pitch in the founder’s voice — she’s blunt and gets to the point.”
• “Add an anti-pattern about candidates who’ve only worked at $1B+ companies.”

Each pass takes 30 seconds. Three passes get you to something genuinely usable.

Because Claude auto-invokes the right skill from what you ask, you can run the whole loop in one thread:

1. “Help me build a brief for our first VP Sales.” → role-intake-brief.
2. “Now turn it into an ICP.” → ideal-candidate-profile.
3. “Build a ranked sourcing search.” → sourcing-search-builder.
4. “Draft a first-touch InMail for the top three.” → outreach-first-touch.

Each skill reads the conversation context: the brief feeds the ICP, the ICP feeds the sourcing string, the sourcing notes feed the outreach. No copy-paste between skills.

Skills hold the procedure. Claude Projects hold your company. If you find yourself pasting the same context into every conversation — your EVP, your comp bands, the way your CEO writes — put it in a Project once and stop.

1. Create a new Project (e.g. “Hiring loop”).
2. Add your standing context to the Project knowledge.
3. Start every hiring conversation inside that Project.

Skills you’ve enabled stay available across all your Projects. Now every conversation starts with both the procedure (skill) and the company-specific context (project) loaded.

A prompt is one instruction. A skill is a procedure — with checkpoints, anti-patterns, a quality bar, and a description that tells Claude when to use it. The practical difference shows up on long, structured tasks: a role brief, a competency scorecard, a multi-touch sequence. Prompts get vague fast; skills hold their shape.

Skills also scale across a team. Once your talent acquisition org agrees on what a good role brief looks like, that agreement lives in the file — and Claude invokes it automatically without anyone needing to remember to. New joiners pick it up the same week they start.