Builds
Things I’ve shipped that close the gap between a connected tool and work that gets done.
hubspot-claude
Open sourceProblemOps teams spend hours copying deals, contacts and tickets out of HubSpot into spreadsheets and chat threads, cleaning them up, then pasting the results back. The stock connection lets Claude read records, but not work across them the way an operator does, so the re-keying never stops and nobody owns the errors it introduces.
What I builtAn open-source Claude Code plugin for HubSpot. It bundles API access and the everyday ops skills into one install, so the agent can look up, update and reconcile records itself instead of handing the work back to a person. Operators describe the job in plain language; the plugin does the clicking.
Agentic patterns skill
SkillProblemTeams starting with agents pick a pattern because it’s popular, not because it fits. A single prompt gets asked to do what needs a pipeline, or a multi-agent setup gets built for a job one skill could handle. The cost shows up weeks later as rework.
What I builtA skill that reads a product requirements doc and recommends which agentic coding patterns fit the use case, with the reasoning for each. It’s published in the Encyclopedia of Agentic Coding Patterns, so teams can check every recommendation against the source.
CRM hygiene agent
AgentProblemDuplicate contacts, stale deals and half-filled records pile up faster than anyone can clean them. Reports drift, reps stop trusting the data, and cleanup turns into a quarterly fire drill that one person dreads.
What I builtA long-running agent that works through the CRM on a schedule: it finds duplicates, flags stale records and fills gaps it can verify, then leaves a short summary of what it changed. Cleanup happens in the background instead of as a quarterly project.
Need something like this for your stack?
Tell me the tools and the job. PromptMetrics scopes and builds it.