Documentation Index
Fetch the complete documentation index at: https://docs.selftune.dev/llms.txt
Use this file to discover all available pages before exploring further.
Overview
The Skills API returns performance data for all skills tracked in your SelfTune workspace, including trigger rates, confidence metrics, and grading health.List skills
Query parameters
| Parameter | Type | Description |
|---|---|---|
page | integer | Page number (default: 1) |
per_page | integer | Results per page (default: 20) |
Response
Skill fields
| Field | Type | Description |
|---|---|---|
skill_id | string | Unique skill identifier |
status | string | healthy, warning, critical, or unknown |
unique_sessions | integer | Number of distinct sessions where the skill was seen |
has_evidence | boolean | Whether graded evidence exists for this skill |
trigger_rate | number | null | Fraction of sessions where the skill fired. null if not enough data. |
avg_confidence | number | null | Average routing confidence (0–1) across all triggers. null if no confidence data. |
confidence_coverage | number | null | Fraction of triggers that include a confidence score. null if not applicable. |
Get skill detail
Response
skill object includes the same trigger_rate, avg_confidence, and confidence_coverage fields as the list endpoint.
Confidence metrics
Three fields describe how confidently and consistently the skill is being routed:trigger_rate — the fraction of recent sessions where the skill fired at least once. A low trigger rate on an active skill may indicate a routing problem or an overly narrow description.
avg_confidence — the mean confidence score assigned by the routing system when it selected this skill. Higher values mean the skill’s trigger description is unambiguous. Values below ~0.6 suggest the skill may be competing with others or have an unclear scope.
confidence_coverage — the fraction of triggers that included a confidence score. Some platforms do not emit confidence data; this field tells you how much of your avg_confidence value is based on actual measurements versus inferred.
All three fields are null when there is not enough data to compute a meaningful value (for example, a newly added skill with no recorded sessions).