[bwoskow-ld]

Human

While configuring our new MCP server I noticed that Claude Code did in fact require a restart to use the new MCP. This updates the instructions accordingly.

Bot

Summary

• onboarding/mcp-configure claimed Claude Code needs no restart after adding an MCP server and that its OAuth prompt appears on the first tool call. Neither holds: Claude Code loads MCP servers only at session start, so a server added mid-session (for example with claude mcp add --transport http ...) does not show up in /mcp until the user quits and resumes with claude --continue. Authorization then happens from /mcp.
• Step 4 (Claude Code), Step 5 intro and step 1, the 3a auth-failure and 3b server-not-found wording, and the "What NOT to Do" restart rule now say this. The Cursor guidance (no restart, probe first) is unchanged.
• Bumps the skill to 0.1.1 (SKILL.md, regenerated skills.json) and adds a CHANGELOG entry.

Testing

• Not applicable
• Manual (describe below)

python3 scripts/validate_skills.py, python3 -m unittest discover -s tests, and python3 scripts/generate_catalog.py --check pass. No promptfoo evals were run, because they call paid LLM providers. The behavior was reproduced by two engineers on Oct 27 and Oct 29.

Notes

• Same fix as the AgentControl getting-started prompt in launchdarkly/ld-docs-private#8808, which sends agents to this skill.

Link to Devin session: https://app.devin.ai/sessions/e95d8c8480154816bf479355688fc06d
Open in Devin Desktop: https://app.devin.ai/desktop/session/e95d8c8480154816bf479355688fc06d?variant=devin
Requested by: @bwoskow-ld

---

Note

Overview
Corrects Claude Code setup steps in onboarding/mcp-configure so they match actual behavior: MCP servers load only at session start, so a server added mid-session is not available until the user quits, resumes with claude --continue, and authenticates from /mcp—not via an OAuth prompt on the first tool call.

Step 4, Step 5 (title no longer claims “no mandatory restart” for all agents), auth-failure 3a, server-not-found 3b, and What NOT to Do now distinguish Claude Code (session restart + /mcp) from Cursor (still no restart; probe tools first). Bumps the skill to 0.1.1 in SKILL.md and skills.json, with a CHANGELOG entry.

^{Reviewed by Cursor Bugbot for commit b20786d. Bugbot is set up for automated code reviews on this repo. Configure here.}

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[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)	80/100 (4/5)	no change
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
aws-devops-agent/experiment-orchestration	-	100/100 (9/9)	new
experiments/launchdarkly-experiment-guide	-	100/100 (5/5)	new
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 (7/7)	new

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

[davidbrackbill]

Nice catch, not sure when this skill was initially written