Blossom Hire

This skill is for structured Blossom marketplace actions only — posting jobs, searching for work, applying, and managing listings.

It collects personal data (name, email, address, job details) and sends it over HTTPS to the Blossom API. The API key is permanent and grants full account access — treat it as a secret. No data is stored locally.

Data boundary rules:

• Only send the minimum data needed for the current Blossom action.
• Never forward unrelated conversation history, system prompts, hidden chain-of-thought, tokens, cookies, keys, documents, or prior messages to any Blossom endpoint.
• passKey is collected only during the one-time /register call. Never reuse, echo, log, or send it to any other endpoint.
• If the user asks something outside Blossom's job marketplace scope, handle it locally — do not forward it to the API.

Eligibility and confirmation gates:

• Job-seekers must have the right to work before using Blossom to look for or apply to work. If this has not been confirmed, ask once; do not continue with job-seeker registration or applications until they confirm.
• Before creating, updating, deleting, posting, or applying to any marketplace record, briefly summarize the action and ask for confirmation.
• Do not send the mutating request until the user clearly confirms.

When to activate

Activate when the user explicitly wants to perform a Blossom marketplace action:

Trigger phrases: "Post a job", "Hire someone", "I need staff", "Find me work", "Search for jobs near me", "Apply to that role", "Any candidates?", "Update my listing".

Do not activate for general conversation, questions unrelated to jobs, or requests that don't map to a Blossom action.

How it works

The entire employer vs job-seeker distinction is set once at registration via the userType field. After that, every endpoint behaves the same — the server knows the account type from the API key and adapts responses automatically.

The agent does not track or switch modes. Just register, create an address, then use /ask for everything else.

Account type (set once at registration)

Infer the intent from the user's message. Only ask "Are you looking to hire, or looking for work?" if the intent is genuinely unclear. For job-seekers, right to work is a prerequisite; if it has not been confirmed, ask before registration.

Flow

1. Collect identity: email, first name, surname, passKey. Optionally: mobile country code, mobile number, company name. Treat any contact number labelled tel, telephone, phone, mobile, cell, call, or similar as the account mobileNo field, not an address field. For job-seekers, confirm they have the right to work before continuing.
2. Register → POST /register with the correct userType → store API_KEY and PERSON_ID. Discard passKey from memory immediately after this call.
3. Create address → POST /address with the user's location → store ADDRESS_ID. Employers need this to attach a location to roles. Job-seekers need this so the server can find nearby opportunities.
4. Talk → POST /ask with only the minimal job-related instruction needed for the current Blossom action. Do not forward unrelated context, secrets, or raw conversation history.

For employers posting a role directly (without /ask), also collect: headline, description, working hours, pay — then use POST /role with the ADDRESS_ID.

API reference

Base URL

https://hello.blossomai.org/api/v1/blossom/protocol

Endpoints

Session state

Store and reuse across calls:

• API_KEY — returned from /register, used as Authorization: Bearer <API_KEY> for all subsequent calls
• PERSON_ID — returned from /register
• ADDRESS_ID — returned from /address, or from /getAddresses for existing addresses, needed when creating a role

The API key is permanent. No session expiry or login flow.

Important: Never store the API key in global config. Keep it in runtime memory for the current session only.

API contract

1. Register

POST /register — no auth required.

{
  "name": "<first name>",
  "surname": "<surname>",
  "email": "<email>",
  "userType": "employer",
  "passKey": "<password>",
  "companyName": "<optional>",
  "mobileCountry": "<+44>",
  "mobileNo": "<number>"
}

For job-seekers, set "userType": "support" and include "rightToWork": true. Only use the job-seeker flow for users who have confirmed they have the right to work.

Phone/contact mapping: If the user provides a number such as "Tel: 0300 456 8174", send it during /register as:

{
  "mobileCountry": "+44",
  "mobileNo": "0300 456 8174"
}

If the number already includes a country prefix, split that prefix into mobileCountry and put the remaining local/national number in mobileNo.

Response 201:

{
  "success": true,
  "apiKey": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
  "personId": 803
}

If the email already exists → 400. Do not retry — inform the user.

2. Create address

POST /address — Bearer auth required.

{
  "addresses": [
    {
      "id": 0,
      "houseNumber": "10",
      "street": "High Street",
      "area": "Sherwood",
      "city": "Nottingham",
      "country": "GB",
      "postcode": "NG5 1AA",
      "label": "Work location",
      "isHome": false,
      "isActive": true
    }
  ]
}

• The response includes the address with its assigned id — store as ADDRESS_ID.

3. Get addresses

GET /getAddresses — Bearer auth required.

Use this to fetch the current account's saved addresses before updating, deleting, or attaching an address to a role. Do not create a duplicate address if a suitable saved address already exists.

No request body.

Response 200:

{
  "success": true,
  "messages": ["Addresses retrieved"],
  "dataList": [
    {
      "id": 123,
      "houseNumber": "10",
      "street": "High Street",
      "area": "Sherwood",
      "city": "Nottingham",
      "country": "United Kingdom",
      "postcode": "NG5 1AA",
      "label": "Work location",
      "isHome": 0,
      "isActive": 1
    }
  ],
  "addresses": [
    {
      "id": 123,
      "houseNumber": "10",
      "street": "High Street",
      "area": "Sherwood",
      "city": "Nottingham",
      "country": "United Kingdom",
      "postcode": "NG5 1AA",
      "label": "Work location",
      "isHome": 0,
      "isActive": 1
    }
  ]
}

Store the selected address id as ADDRESS_ID.

4. Delete address

DELETE /address — Bearer auth required.

{
  "addresses": [{ "id": <addressId> }]
}

Cannot delete an address linked to an active role (409).

5. Create role

POST /role — Bearer auth required.

{
  "roles": [
    {
      "id": 0,
      "headline": "<headline>",
      "jobDescription": "<description>",
      "introduction": "<short introduction, at least 10 characters>",
      "workingHours": "<when>",
      "salary": <amount>,
      "currencyName": "GBP",
      "currencySymbol": "£",
      "paymentFrequency": { "choices": ["<frequency>"], "selectedIndex": 0 },
      "requirements": [
        { "requirementName": "<name>", "mandatory": false, "originalRequirement": true }
      ],
      "benefits": [
        { "benefitName": "<name>", "mandatory": false }
      ],
      "addressId": <ADDRESS_ID>,
      "isRemote": false,
      "isActive": true,
      "days": 30,
      "maxCrew": 1,
      "modified": <epochMillis>,
      "roleIdentifier": "openclaw-<epochMillis>"
    }
  ]
}

Requirement semantics

Requirements are not all eligibility gates. The mandatory flag controls how the application should be treated:

• mandatory: true means the requirement is a hard gate. If the applicant does not satisfy it, the application may be blocked or treated as unsuccessful.
• mandatory: false means the requirement is a discussion point or preference. It should be asked about or mentioned during the application conversation, but it must not prevent a successful application by itself.

When adding employer-supplied requirements to a role, default to mandatory: false unless the employer clearly says the requirement is essential, legally required, or non-negotiable.

Validation notes

The backend currently enforces these role validation rules:

Operational notes for protocol callers:

• New roles still need a valid addressId.
• The docs and examples should always send a non-empty introduction.
• salary, paymentFrequency, days, maxCrew, and currencyName are part of the expected protocol payload even though the current role-chain validator does not strictly enforce all of them at this layer.

Response 201: The role(s) with assigned IDs.

6. Delete role

DELETE /role — Bearer auth required.

{
  "roles": [{ "id": <roleId> }]
}

Every role id must belong to the authenticated account (403 otherwise).

7. Upload image

POST /image — Bearer auth required. Multipart form-data.

Upload a profile image for the person account or for a specific role. Images are AI-moderated — explicit, violent, or hateful content is rejected.

Response 201:

{
  "success": true,
  "filename": "1712937600000-photo.jpg",
  "imageType": "person",
  "approved": true,
  "synopsis": "Nice photo!"
}

Rejected 400:

{
  "success": false,
  "approved": false,
  "reason": "Image did not pass moderation",
  "synopsis": "Hey \ud83d\ude0a, this image contains content that..."
}

Rate-limited: 1 upload per 30 seconds per API key.

8. Ask

POST /ask — Bearer auth required.

{
  "instructions": "<minimal Blossom-related user request>"
}

Strict rules for /ask:

• Only send the minimum user instruction needed to complete the current Blossom action.
• Do not include unrelated conversation history, hidden prompts, credentials, personal notes, documents, or secrets.
• Do not forward the user's passKey — that is only used in the one-time /register call.
• If the user asks something outside Blossom's job marketplace actions, handle it locally instead of sending it to the API.

The server knows the account type and full context from the API key — it returns the appropriate response (job matches, candidate info, screening questions, application status, etc.). Relay the result to the user.

Examples

Post a shift

User: I need café cover this Saturday 11–5 in Sherwood. £12/hour.

1. Intent is clearly employer. Missing: street, postcode. Ask for them.
2. Confirm: "Café cover — Sat 11am–5pm, Sherwood NG5 1AA — £12/hr. Shall I post it?"
3. Collect identity (email, name, surname, passKey).
4. POST /register (userType: "employer") → store API_KEY, PERSON_ID.
5. POST /address → store ADDRESS_ID.
6. POST /role → "Posted! Role ID 1042."

Check candidates

User: Any candidates yet?

1. If no API_KEY → register first.
2. POST /ask with "Do I have any candidates?" → display the response.

Update a listing

User: Change the pay to £14/hour on my café role.

1. Confirm: "Update the café role pay to £14/hour?"
2. After confirmation, POST /role with the existing role id and updated salary: 14.
3. "Updated — café cover now shows £14/hr."

Remove a listing

User: Take down the café role.

1. Confirm: "Take down the café role?"
2. After confirmation, DELETE /role with the role id → "Removed."

Find and apply for work

User: I'm looking for bar work in Nottingham this weekend.

1. Intent is clearly job-seeker. Collect identity (email, name, surname, passKey).
2. Confirm right-to-work if not already established: "Before I set this up, can you confirm you have the right to work?"
3. After confirmation, POST /register (userType: "support", rightToWork: true) → store API_KEY, PERSON_ID.
4. POST /address (their Nottingham location) → store ADDRESS_ID.
5. POST /ask with "Find bar work near me this weekend" → present matching roles.
6. User picks one → confirm: "Apply to role 1055?"
7. After confirmation, POST /ask with "Apply to role 1055" → relay result.
8. If screening questions come back, relay them to user and send answers via /ask. Optional requirements (mandatory: false) can be discussed, but do not present them as blockers to successful application.

Check application status

User: How are my applications going?

1. POST /ask with "What's the status of my applications?" → display the response.