[sattensil]

Summary

• Initialize with init_client / get_client; the complete workflow is async and ends with await shutdown().
• New section on attributing custom metrics to a config variation with result.track_data.

Jira: DEVREL-1649

Testing

• Not applicable
• Manual (describe below)

Ran the complete workflow in custom-metrics.ipynb; the new metric appears in the metric list.

Notes

• Targets launchdarkly-ai-server 0.3.0 (Python) and @launchdarkly/ai-node 0.3.0 (Node.js).
• Part of a set of six roll-forward PRs (migrate, built-in-metrics, projects, tools, online-evals, custom-metrics). Each regenerates skills.json, so later merges may need python3 scripts/generate_catalog.py re-run.
• The matching notebook in launchdarkly-labs/agentcontrol-cookbooks mirrors this skill's code; that update lands separately.

---

Note

Overview
Updates the custom-metrics agent skill (and matching skills.json entry to 1.1.0-experimental) so metric lifecycle guidance matches the new LaunchDarkly AI SDK and hosted MCP.

Create and read paths now lead with MCP (create-metric, get-metric, list-metrics, list-metric-events); update/delete stay on the REST API with a writer token. SDK examples switch from raw ldclient setup to launchdarkly-ai-server / @launchdarkly/ai-node (init_client, get_client, await shutdown() instead of flush()), with LD_SDK_KEY and prerequisites pointing at sdk-install.

Adds documentation for attributing custom metrics to config variations by passing track_data / trackData from config().invoke() into track(), plus an MCP-first create payload example. Compatibility text and doc links now reference the AI SDK references instead of the legacy Python SDK docs.

^{Reviewed by Cursor Bugbot for commit f418c1e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Skill eval results

Skill	Before	After	Δ
agentcontrol/configs-create	100/100 (4/4)	100/100 (4/4)	no change
agentcontrol/configs-update	80/100 (4/5)	100/100 (5/5)	+20
agentcontrol/configs-variations	80/100 (4/5)	80/100 (4/5)	no change
agentcontrol/tools	75/100 (3/4)	75/100 (3/4)	no change
feature-flags/flag-and-release-change	-	100/100 (4/4)	new
feature-flags/flag-release	-	80/100 (4/5)	new
feature-flags/launchdarkly-flag-command	-	100/100 (3/3)	new
feature-flags/launchdarkly-flag-create	100/100 (3/3)	100/100 (4/4)	no change
feature-flags/launchdarkly-flag-drift	-	100/100 (4/4)	new
feature-flags/should-flag-change	-	100/100 (17/17)	new
onboarding	-	100/100 (4/4)	new

Only suites whose source actually changed since their last recorded score were re-run. Soft-failing while we stabilise the baseline.

[jeffdupont]

Approving with one nit. Nothing PR-specific found; only the cross-cutting install note applies.

---

skills/agentcontrol/custom-metrics/SKILL.md:17 — Add launchdarkly-server-sdk to the Python prerequisites. The snippets import ldclient.Context, and launchdarkly-ai-server doesn't install it.