Skip to main content
GET
Python SDK

Authorizations

Authorization
string
header
required

Enter your API key (e.g. dai_sk_test_key_001)

Headers

Authorization
string

Query Parameters

q
string | null

Search query (keyword or semantic)

category
string | null

Filter by category

tags
string | null

Comma-separated tags to filter by

badge
string | null

Filter by badge — where the skill came from, not a quality tier: community (authored on the platform), imported (synced from GitHub), featured (auto-promoted), verified (retired tier, no skill carries it)

skill_type
string | null

Filter by taxonomy class: capability, preference (legacy model-gap/proprietary/convention also accepted; unlabeled skills only surface without the filter)

invocation
string | null

Filter by invocation mode: model, user, any

sort
string
default:recommended

Sort by: recommended (SkillScore v2, evidence-tiered, ≥10 activations; alias: effectiveness), lift (verified benchmark lift; alias: biggest_improvement), popular (30d uses, installs tiebreak), installs, top_rated (AI live rating, ≥10 rated), efficiency (token savings), most_turn_savings, rating (legacy pass-rate), recent

cursor
string | null

Opaque pagination cursor — pass back the next_cursor from the previous page. A cursor is tied to the filter set and sort it was issued under; change either and paging restarts at page 1. It does NOT expire and is not tied to a session, so a client may store one and resume the same walk on a later request. Stop when has_next is false. A truncated: true field means the walk reached the depth limit for this ordering rather than the end of the results — re-issue under sort=recommended, lift or recent, which page to any depth

limit
integer
default:20

Results per page

Required range: 1 <= x <= 100
view
string
default:grid

Display view — 'grid' (default) or 'ranks' (Leaderboard: adds rank + rank_metric fields)

include_unsafe
boolean
default:false

Platform-admin only: include skills that did not pass safety review. Ignored for all other callers (default false).

measured
string
default:default

Measurement facet: 'default' (all skills, measured ranked first), 'only' (curated shelf — measured imports only), 'all' (alias of default)

embedding_space
string | null

Diagnostic: rank against this embedding space instead of the live one. Exists so the cutover gate can score a building space BEFORE it is flipped live — without it the gate has nothing to read until after the flip it is meant to authorise. Unknown/retired ids fall back to live.

search_mode
string | null

Diagnostic: 'ilike' forces the keyword ranker for this request only. Anything else is ignored. Exists so both rankers can be scored against the same corpus in the same minute when evaluating a search-ranking change.

Cookies

decimal_session
string | null

Response

Successful Response