[FEATURE] Tool_result event skill telemetry/metrics not captured for subsequent invocations in the same session

Resolved 💬 5 comments Opened Mar 19, 2026 by egplat Closed Apr 17, 2026

Preflight Checklist

  • [x] I have searched existing requests and this feature hasn't been requested yet
  • [x] This is a single feature request (not multiple features)

Problem Statement

Issue Title

Skill telemetry/metrics not captured for subsequent invocations in the same session

## Description
When a skill is invoked multiple times within the same Claude Code session, the skill loading prompt ("Launching skill: ...") only appears on
the first invocation. Subsequent invocations reuse the loaded skill context without triggering the loading event again.

If OpenTelemetry (otel) metrics or other telemetry data are only emitted during skill loading, this leads to significant underreporting of
actual skill usage
, as only the first invocation per session is counted.

## Current Behavior

  1. First invocation: Skill loads → "Launching skill: xxx" appears → Telemetry captured ✅
  2. Subsequent invocations: Skill executes directly → No loading event → Telemetry NOT captured ❌

## Expected Behavior
Every skill invocation should be tracked in telemetry/metrics, regardless of whether it's the first call or a repeated call within the same
session.

## Impact

  • Underestimated PV (Page View) / usage metrics: Real usage frequency is much higher than reported
  • Inaccurate skill analytics: Cannot properly measure skill engagement or popularity
  • Misleading data for skill developers: Makes it hard to understand actual user behavior patterns

Proposed Solution

Every skill invocation should be tracked in telemetry/metrics, regardless of whether it's the first call or a repeated call within the same session.

Alternative Solutions

_No response_

Priority

Critical - Blocking my work

Feature Category

Other

Use Case Example

_No response_

Additional Context

_No response_

View original on GitHub ↗

This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗