Review: install + trigger test across AI systems (v0.1.0)

[shah-siddd]

Pre-public review of the openlayer skill (v0.1.0). Please install it in your AI coding tool and confirm it (a) installs, (b) loads/triggers, and (c) produces correct guidance. Comment with pass/fail + what the agent did.

Access: repo is private — you need read access. For the skills CLI you also need a GitHub token in your environment.

A. Claude Code (primary)

• claude plugin marketplace add openlayer-ai/openlayer-skills
• claude plugin install openlayer@openlayer
• claude plugin details openlayer@openlayer shows 1 skill
• In a fresh session on a small app, prompt: "add Openlayer tracing to this app" → the agent loads the openlayer skill, sets OPENLAYER_API_KEY + OPENLAYER_INFERENCE_PIPELINE_ID, wraps the client (trace_openai/etc.) + @trace, and does not hardcode keys.

B. skills CLI (Cursor / Windsurf / others)

• npx skills add openlayer-ai/openlayer-skills --skill "openlayer"
• Skill appears in the agent's skills list
• Same trigger prompt as above → correct instrumentation

C. Manual (fallback / any agent)

• Clone + ln -s <repo>/skills/openlayer <agent-skills-dir>/openlayer
• Skill loads + triggers

Also spot-check a dev-mode prompt

• "set up Openlayer offline tests for this project" → authors valid openlayer.json + tests.json using real catalog subtypes (not invented), runs openlayer validate.

Report

For each system: ✅/❌ install, ✅/❌ trigger, and a sentence on the agent's behavior. File any wrong/outdated guidance as a separate issue (template: Skill issue or improvement).

Once this checks out across systems, we flip the repo public and submit the Cursor plugin.