Recruiter for Instinct: build and manage hiring with Badal

A recruiter using Instinct can use Badal as the hiring system when the agent has an authenticated way to operate it. Badal provides public application forms, private candidate material, versioned screening criteria, assessments, candidate reports, and funnel analytics. Instinct can coordinate the work through available MCP, HTTP, or browser tools. No native Instinct installation is claimed by this guide.

The important distinction is between understanding the service and accessing an account. Any web-reading agent can discover Badal’s documentation. Reading a page cannot grant permission to create a job, inspect a resume, charge a card, or reject a candidate. This guide describes the service contract and a practical workflow, rather than relying on hidden page instructions.

Give the agent a factual starting point

Badal’s capability manifest lists the transports, authentication requirements, account verification endpoint, recruiting setup fields, and actions that need approval. The service guide explains product behavior. The docs describe integrations. OpenAPI supplies exact request schemas.

These resources allow an agent to discover the same facts a person can read. They contain no private account information. The recruiter’s request and the host’s instructions determine what the agent is authorized to do.

Before configuring the role, have the agent identify its actual execution path:

Do not assume that web browsing supports private headers or that browser cookies transfer to a different tool. If verification fails, explain the missing capability and stop account mutations. The recruiter can still use the draft in the workspace.

Connect privately and limit unnecessary access

The recruiter creates a manual agent key or a one-time connection link in Agent access. Store credentials in the host’s private credential store. A local environment that supports the Badal Recruiter skill can use its helper; a text-only chat cannot gain HTTP execution by downloading instructions.

The temporary connection flow uses a one-time redemption link and time-limited access. Manual keys remain valid until revoked. These are different mechanisms: do not promise that a manual key expires after 24 hours. Revoke access when changing agent environments or ending a project.

Keys never belong in job-post copy, application fields, candidate messages, public URLs, or logs. An admin-dashboard key is also different from a recruiter agent key and must not be used for the recruiting workflow.

Use a progressive intake for the recruiter

A good setup conversation asks one missing detail at a time. Start with the organization and recruiter name candidates should see, then the role title. Ask what success looks like in the first month. Clarify duties and work arrangement before choosing screening criteria.

Next decide what evidence the application should collect. A resume can provide background, but relevant work material and a clear explanation of personal contribution often make a report more useful. Confirm which fields are required. Then agree on three to six job-related criteria, their weights, and their minimum scores.

The last setup topics are the exercise, permitted tools, email context, interview booking link, and Google Sheets. Show the completed draft before publication. Avoid inventing recruiter details merely to complete an API request.

Example: full stack intern

Suppose the recruiter wants an intern who can implement a small API and explain their work. Suggested application material includes a resume, one project link, a description of the candidate’s contribution, and an example of debugging a failure.

A starting rubric could cover implementation fundamentals, debugging, data handling, and technical communication. The recruiter should decide weights and thresholds based on the actual duties. Engineering education can be relevant context, but it should not replace evidence of the skills the job requires.

A 45–60 minute assessment could ask the candidate to design and implement a small endpoint with validation, a simple persistence layer, and one meaningful test. Supply the input shape, expected output, allowed tools, and evaluation criteria. Permit a written explanation when a candidate cannot use the preferred runtime. This reduces accidental testing of environment setup rather than the job’s core skill.

The agent’s review should point to the validation behavior, handling of errors, test quality, and explanation of tradeoffs. It should identify uncertainty rather than treating an untested repository as proven implementation quality.

Example: GTM or content and operations intern

For a GTM role, request a campaign or writing sample and an explanation of the audience, goal, and result. Ask the applicant to reason about a small experiment: channel choice, creative hypotheses, measurement, and conditions for stopping or iterating.

An assessment might provide a product brief and ask for an ad, a landing-page hero, and a short rationale. Include the customer segment and one concrete pain point. A useful rubric distinguishes clarity, audience insight, message quality, and testing reasoning.

For operations, give a fictional support backlog or workflow with missing information. Ask the candidate to prioritize, identify assumptions, and propose a simple follow-up system. Evaluate how they reduce ambiguity and communicate decisions. More examples are available in the role playbooks and Muse recruiter guide.

Manage the pipeline through evidence and status

After publishing, return the real application link. Applicants complete a one-question-per-screen form. The agent can list candidates, open detailed reports, and explain the rubric version used for each evaluation.

A useful daily review has three parts: new applicants, candidates waiting on an action, and processing failures. Assessment sent, assessment submitted, score unavailable, manual review, and interview invited represent different states. Do not collapse them into a vague “passed” label.

Read delivery and processing jobs when an email or evaluation appears delayed. A malformed file can require manual review of the original material. Do not repeatedly retry a failed action without first checking whether the earlier attempt succeeded.

The recruiter approves final keep or reject decisions for the specific candidate. MCP final decisions additionally require approval in Badal’s signed-in candidate report. The external booking service manages slots; Badal sends its link but does not confirm the booking.

Keep collection open when credits run out

Badal’s public forms and Stage 1 Google Sheets synchronization remain available beyond the free AI screening allowance. Recruiters who prefer manual review can keep collecting applications and material. After adding credits, they can authorize processing of waiting candidates.

The current packs are $5 for 25 candidates, $50 for 278, and $500 for 3,334. Larger quantities round up, making the effective rates approximately $0.18 and $0.15 per candidate. Read GET /api/billing for the active offer instead of assuming a blog price is permanent.

Auto-reload charges a saved card only after explicit approval of the selected pack. Configuration is not itself a payment. Credits are added after verified Stripe success. A payment that fails or requires authentication needs human attention, not repeated attempts to charge the recruiter.

Use analytics for decisions about the process

Badal exposes recruiting analytics through the dashboard, HTTP API, and MCP. Compare role-level stages and counts to find where applicants are waiting. Use this to investigate the process: an unclear assessment, missing booking link, unavailable credits, or a failed job may explain a bottleneck.

Counts alone do not establish candidate quality or prove that an interview occurred. Small samples also make conversion percentages unstable. Pair stage metrics with a review of actual evidence and operational status before changing the rubric. Earlier reports retain their evaluation version when criteria change.

What completion should look like

For the first hiring session, expect verified account access, an approved role draft, a published application link, the correct resume requirements, a contextual assessment, and an honest integration status. For ongoing work, expect evidence summaries, readable analytics, clear failure reporting, and explicit requests for final decisions.

If the host cannot execute authenticated tools, keep the work at the draft stage and use the workspace for publication. Badal’s integration documentation is the authoritative setup reference; this article offers examples, not permission to act on a recruiter’s account.