HireData skills for your AI assistant

The HireData MCP connection gives your AI assistant HireData’s tools. Skills give it the working method behind them: how an experienced HireData consultant would approach the task, what to check in your workspace first, and when to stop and show you a preview instead of creating something you didn’t ask for. HireData currently publishes ten skills: one that plans, four that build, two that diagnose, one that reports on what the automations delivered, one that writes issue reports, and one that supplies the staffing domain knowledge behind the rest. The skills that build something gather the available context, show you a complete draft or preview, and only create or change something after you approve that exact action. New triggers stay inactive, forms stay unpublished where drafts are supported, and nothing is sent or activated as a side effect. The skills that diagnose read your workspace data, explain what actually happened, and propose a repair. They never apply it on their own.

​

Plan an activation

Turns a recruitment or staffing problem into a small, prioritized activation plan. The skill establishes your operating context (ATS, staffing model, audiences, channels, current HireData usage) and expresses each problem as a user story. It scores opportunities by reach, frequency, value, data readiness, effort, and risk. It then recommends the smallest coherent first release, normally three to five workflows. Each workflow includes a trigger moment, channel, owner, success measure, and a safe first test. Try:

• “We lose candidates between application and first interview. Use HireData to propose which automations to implement first. Do not create anything yet.”
• “Review our existing templates and automations and rank the highest-impact missing workflow.”
• “Build a 30-day activation plan for interview no-show reduction, with success measures.”

When you approve a recommendation, this skill hands the workflow to the matching creation skill below.

​

Create triggers

Designs automation triggers around the business event and the data your later workflow steps need. The skill inspects your connected apps, objects, fields, and existing triggers, checks for overlapping triggers, and previews the source, timing, timezone, filters in plain language, expected frequency, and duplicate controls before creating anything. New and changed triggers stay inactive unless you explicitly ask for activation. Try:

• “Create a trigger for candidates whose placement ends in 30 days. Show me the filters in plain language first.”
• “Our welcome automation fires too often. Diagnose the trigger and propose a narrower configuration.”

​

Create forms

Creates the shortest form that captures the decision or signal your workflow needs, such as screening, availability, interview feedback, intake, consent, or NPS. The skill won’t ask candidates for data HireData already knows from the trigger, person, application, or vacancy. It keeps required fields to a minimum and checks that every branch reaches an appropriate ending. It previews fields, validation, branching, and the submission outcome before creating. Try:

• “Create a candidate availability form for weekend shifts. Show me the form and branching logic before creating it.”
• “Design an interview feedback form for hiring managers that takes under a minute to complete.”

​

Create email templates

Designs emails around the recipient’s context and the next useful action. The skill selects deterministic variables for facts and uses modifiers for formatting and fallbacks. It applies AI variables only for bounded generation with an explicit source, tone, length, and fallback. An AI variable is never allowed to invent dates, salary, legal status, or commitments. You see the complete copy, subject, variables, and CTA destinations before anything is saved. Test emails go only to an address you approve. Try:

• “Design an application acknowledgement email that uses an AI variable safely. Show me a preview and do not send it.”
• “Improve our interview invitation template: check which variables can replace the hard-coded text.”

​

Create WhatsApp templates

Creates concise WhatsApp templates that are useful in context and likely to pass Meta’s current review requirements. The skill reasons explicitly about the Meta category (it won’t disguise promotional content as transactional), validates variable order and button targets, plans the response and no-response paths, and reports Meta approval as pending until the platform confirms it. No live message is sent as part of template creation. Try:

• “Create a WhatsApp template reminding candidates of tomorrow’s interview, with a confirm button and a reschedule path.”
• “Review our shift-offer template: is the category right, and what would you change before resubmitting to Meta?”

​

Diagnose automations

Explains what actually happened to a workflow before anything is changed. Most reports of a “broken” automation turn out to be one of several different situations with different fixes: a trigger that never produced an event, runs that stopped at a filter, a step that errored, a message that was sent but never delivered, runs that started and never finished, or an automation running at a volume nobody intended. The skill separates them using your events, runs, steps, and message data, then states what happened, the evidence for it, how many records it affected, whether it is still happening, and what remains unexplained. It also reports what the volume actually cost. Repairs come with their trade-off, and nothing is changed, paused, activated, or replayed without your approval of that exact change. Try:

• “This candidate never received our welcome email. Use HireData to find out why.”
• “Our onboarding automation shows thousands of cancelled runs. Is that a fault, or normal?”
• “Check whether any of our automations are stuck, or producing far more runs than results.”

​

Diagnose email performance

Separates “the mail didn’t arrive” from “the mail arrived and nobody cared”. Those feel identical when you look at a falling open rate, and they have nothing in common. The skill first checks delivery, bounce, complaint, and unsubscribe rates against healthy thresholds, so you know whether the sending itself is at fault. It then strips machine opens — Apple Mail Privacy Protection inflates every headline open rate — and compares human-only engagement per provider and per month, using click-to-open to tell a content problem apart from an inbox placement one. It also catches comparisons that aren’t comparisons, such as one-to-one mail measured against a bulk campaign, or a previous system that counted opens differently. Advice comes back in order of expected impact, with what HireData will do separated from what you need to do. Try:

• “Our open rate dropped since we moved to HireData. What changed?”
• “Mail to Microsoft addresses seems to be blocked. Check the last six months per provider.”
• “What are our delivery, bounce, and complaint rates this quarter, and are they healthy?”

​

Report automation ROI

Answers “what has this actually given us?” without inventing anything. The skill first establishes the true reporting period from when your automations were created and when they actually ran, rather than accepting a window someone picked. It separates automations that communicate with candidates and contacts — which can be credited with placements — from data-only automations, which are credited with time saved. Every figure comes from measured workspace data, never a model, and where attribution covers less of the period than the automations were live, the skill says so and explains why. You get the customer-facing readout plus an internal note listing the numbers it rejected and the arithmetic that failed. Try:

• “What have our automations delivered since we went live? Use only measured figures.”
• “Build a business review for this workspace ahead of renewal, and tell me which numbers you couldn’t stand behind.”

​

Write issue reports

Turns a bug, customer request, prototype, recording, meeting note, or technical prompt into a structured HireData Linear issue. The skill includes HireData’s issue-writing method, templates, refinement guidance, and maintained workflow snapshots. HireData team members can use the live Linear workspace and internal Notion SOPs; approved partners can use the project and Notion pages shared with them. Users without live access still get a complete report with unverified metadata clearly marked. Installing the skill never grants additional access. Existing HireData, Linear, and Notion permissions determine which sources and live actions are available. Try:

• “Turn these reproduction steps, screenshot, and automation URL into an intake-ready HireData bug issue. Draft only.”
• “Refine this prototype into a product specification, but treat the mock as evidence rather than the final product contract.”
• “Use the HireData Notion pages shared with our partner project to improve this issue, and mark inaccessible metadata as unverified.”

Learn how to write better HireData issue reports with AI.

​

Apply staffing lifecycle knowledge

Background knowledge rather than a task you ask for directly. It gives your assistant staffing and ATS domain judgment: who the actors are (candidate, contact, company, recruiter) and who owns each record; how candidates, jobs, matches, publications, talent pools, and placements relate; which stage a match is in and the event that legitimately moves it forward; and what a candidate-driven or job-driven market means for timing, audience, and cadence. Mostly it prevents expensive misreadings — “close the candidate” after an assignment ends means ending the placement, not the person, and a rejected match cannot trigger onboarding. The other skills work without it; it makes them read your request correctly. It supplies reasoning, never workspace facts: stage names, fields, and values are always confirmed against your live workspace. Try:

• “This placement ends in three weeks. What should happen next for this candidate?”
• “The client hasn’t given feedback on our two submissions. What should we send, to whom, and when?”

​

Get the skills

The skills are bundled with the HireData MCP connection in the HireData plugin. One install covers both. Install the HireData plugin in ChatGPT, Codex, or Claude. No plugin support on your plan or assistant? Install the skills with a prompt instead. HireData improves these skills as the product changes and adds new ones, so keep them up to date. They are public and open to review at github.com/HireData/skills, and you can create your own for the work only your team does.