Skip to content
Back to skills

Iblai Api Search

ASecurity

Discover agents and learning content in an ibl.ai organization via the platform API — faceted, paginated search over agents and the catalog (courses, programs, pathways, skills), personalized (RAG) recommendations, global/personalized agent search, and seller-facing sellable-items search; plus admin management of the recommendation prompts. Use to find, browse, or get recommended agents/content (mostly read-only; recommendation-prompt endpoints write).

  • 15 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 20, 2026
ai-agentsgobashapi

Works with

  • api

Security analysis

A100/100

Scanned September 24, 2026

npx -y skills add iblai/vibe --skill iblai-api-search --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Iblai Api Search?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Iblai Api Search
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/iblai-iblai-api-search/badge)](https://www.skillsdirectory.com/skills/iblai-iblai-api-search)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: iblai-api-search
description: Discover agents and learning content in an ibl.ai organization via the platform API — faceted, paginated search over agents and the catalog (courses, programs, pathways, skills), personalized (RAG) recommendations, global/personalized agent search, and seller-facing sellable-items search; plus admin management of the recommendation prompts. Use to find, browse, or get recommended agents/content (mostly read-only; recommendation-prompt endpoints write).
metadata:
  kind: api
---

# iblai-api-search

Discover agents and learning content from the API in three ways: **agent**
search (faceted, full-text search over agents), **content** search
(faceted search over the catalog of courses, programs, pathways, and skills),
and **recommendations** (personalized, RAG-ranked results for the signed-in
user) — plus a newer `/api/ai-search/...` family (global + personalized agent
search, sellable-items search, and admin recommendation-prompt management).
Discovery is read-only; only the recommendation-prompt endpoints write. To
*edit* an agent use the `/iblai-api-agent-*` skills.

## Auth & conventions

- **Base URL:** `https://api.iblai.app/dm` — these are Data Manager (DM)
  endpoints, so the **`/dm` prefix is required**; the paths below are appended
  to it (e.g. `https://api.iblai.app/dm/api/search/catalog/`).
- **Header:** `Authorization: Api-Token $IBLAI_API_KEY` on every request.
- **Path vars:** `{org}` = `$IBLAI_ORG`, `{username}` = `$IBLAI_USERNAME` (agent
  search only; the `/api/ai-search/...` v2 endpoints and content search are
  **not** org-pathed — the org/user come from the token or query params).
- Not connected yet? Run **`/iblai-api-login`** first to populate `IBLAI_ORG`,
  `IBLAI_USERNAME`, and `IBLAI_API_KEY`.

## Reads

### Agent search

- **GET** `https://api.iblai.app/dm/api/search/orgs/{org}/users/{username}/mentors/` — search/browse agents. Query params:
  - `query` — full-text over name/description (e.g. `calculus`).
  - `category` — numeric category id (from `facets.Category`).
  - `created_by` — author username (from `facets["Created By"]`).
  - `limit`, `page` — pagination.

**Envelope:** `{ results[], count, next, previous, current_page, total_pages, facets }`.
Each result has `unique_id`, `name`, `description`, `llm_provider`, `llm_name`,
`categories`, `created_by`, `is_featured`, `mentor_visibility`, `slug`,
`profile_image`, `settings`. `facets` keys: Category, Created By, Featured, LLM
Provider, Subject, Audience, Promotion, Recently Accessed.

### Content search

- **GET** `https://api.iblai.app/dm/api/search/global/` - search/browse the
  catalog (`/api/search/catalog/` is the older spelling of the same endpoint).
  Query params:
  - `query` — full-text (e.g. `manufacturing`).
  - `limit`, `page` — pagination.
  - Filter params mirror the response `facets`: content
    **type** (course / program / pathway / skill), **language**, **level**,
    **subject**, **format**, **price**, **certificate**.
  - `price` - `free` or `paid`. `free` means every course not explicitly marked
    paid, including those with no audit setting recorded, so `free` and `paid`
    together cover the catalog and match the `price` facet counts.
  - `self_paced` - `self-paced` or `instructor-led`; repeat the param to ask for
    both. `courseformat` is accepted as the older spelling. Filters on courses
    whose format is recorded as a boolean, which is narrower than the `format`
    facet counts.

**Envelope:** `{ results[], count, next, previous, current_page, total_pages, facets }`,
where each result is `{ "type": "course|program|pathway|skill", "data": { } }`.

### Recommendations

Personalized, RAG-ranked results for the signed-in user. Unlike the two
searches above, there is **no query**: results are computed server-side and are
**user-dependent** (personalized to whoever the token resolves to). The org and
user are derived from the token; this endpoint is not org-pathed.

- **GET** `https://api.iblai.app/dm/api/ai-search/recommendations/` — RAG recommendations. Query params:
  - `recommendation_type` — `mentors` | `courses` | `programs` | `resources` | `pathways`.
  - `limit` — maximum number of recommendations.

Returns a ranked list of the requested type. Call once per type for a mixed set.

A newer `/api/ai-search/...` family (no org/user in the path — context comes from
the token or query params). The `recommendations/` endpoint above is part of it.

### Global agent search

- **GET** `https://api.iblai.app/dm/api/ai-search/mentors/` — global agent search
  across the platform. Works **anonymously**; if a token is sent, results
  are personalized to that user (pass `platform_key` for RBAC). Query params:
  `query`, `platform_key`, facet filters mirroring the response, and
  `limit`/`offset` pagination. Distinct from the org-pathed agent search above
  (that one is organization-scoped; this one is global).

### Personalized agents

- **GET** `https://api.iblai.app/dm/api/ai-search/personalized-mentors/` —
  personalized, agent-focused results for the signed-in user (authentication
  required). Pass `platform_key` (or `tenant`) and, where applicable, `username`.

### Sellable items

- **GET** `https://api.iblai.app/dm/api/ai-search/sellable-items/` — seller-facing
  search over items that can be sold (for paywall/pricing setup). Requires an
  authenticated user **with the `CanSellItems` RBAC permission**. Query params:
  `platform_key`, `query`, item `type`, and pagination.

### Recommendation prompts (admin)

Manage the LLM prompt(s) that drive recommendations. **Platform-admin only.**

- **GET** `https://api.iblai.app/dm/api/ai-search/recommendation/prompts/`
  (preferred alias of `…/prompts/`) — list prompts. Params: `platform_key`,
  `recommendation_type`, `active_only`, `prompt_id`.

## Writes

### Recommendation prompts

Manage the LLM prompt(s) that drive recommendations. **Platform-admin only.**

- **POST** `https://api.iblai.app/dm/api/ai-search/prompts/` — create a prompt. Confirm with the user first.
- **PUT** `https://api.iblai.app/dm/api/ai-search/prompts/` — update a prompt (by `prompt_id`). Confirm with the user first.
- **DELETE** `https://api.iblai.app/dm/api/ai-search/prompts/` — delete a prompt (by `prompt_id`). Confirm with the user first.

## Example

Find calculus agents, then manufacturing courses:

```bash
curl -s "https://api.iblai.app/dm/api/search/orgs/$IBLAI_ORG/users/$IBLAI_USERNAME/mentors/?query=calculus&limit=10" \
  -H "Authorization: Api-Token $IBLAI_API_KEY"

curl -s "https://api.iblai.app/dm/api/search/global/?query=manufacturing&limit=10" \
  -H "Authorization: Api-Token $IBLAI_API_KEY"
```

## Notes

- Discovery (agent/content/recommendations/global-agent/sellable-items search) is
  read-only; the only writes are the admin **recommendation-prompt** endpoints
  (`POST`/`PUT`/`DELETE …/ai-search/prompts/`).
- The `facets` block in each response enumerates the available filter dimensions
  and counts; use its keys to drive additional filter params.
- An agent result's `unique_id` is the `{mentor}` you pass to the
  `/iblai-api-agent-*` skills; a content result's `data` shape depends on its
  `type`.
- **Search vs recommendations:** search takes an explicit `query` + filters and
  is the same for any caller; recommendations take no query and are personalized
  per user (token-dependent).

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.