Rocky Verified Work Pattern

Rocky is useful when the work has a shape: a request, a target, tools that can act, and a way to verify the result. The pattern below is the public version of that operating loop.

The loop

• State the mission. Convert the user’s request into a clear target and constraints.
• Inspect the source of truth. Read the file, page, API, repo, dashboard, or live system instead of guessing.
• Act with the right tool. Patch files, call APIs, run builds, use the browser, or generate media only after the needed context is known.
• Verify the outcome. Use real output: HTTP checks, screenshots, logs, file reads, API responses, rendered pages, or live URLs.
• Report clearly. Say what changed, where it lives, what was verified, and what is still blocked.

What makes this different from chat

A chatbot can describe a plan. An operator can change the site, post the video, run the test, inspect the output, and return with evidence. The important distinction is not personality — it is closure.

Good targets for the pattern

• Website edits and deployment checks
• Social posts and content queues
• AI news/article curation
• Build documents and PDFs
• Email monitors and routing
• Cron jobs and watchdogs
• Local Mac workflows that need screenshots

Verification examples

rocky@demo:~/workspace$ node --check server.mjs
rocky@demo:~/workspace$ curl -I https://heyrocky.ai/wiki
HTTP/2 200

Public screenshots should use demo paths and fake prompts. Never publish a real home directory, username, token, cookie, or private customer data.