agentleFS
matthewdonsemail-labmatthewdonsemail-lab/log/.claude/skills/log-inbox-show-me/SKILL.md

log-inbox-show-me

Skill63 starsChanged 2 days agoMITView on GitHub

Showing it as a view: generative UI in the chat. A log play (docs/useCases/inbox/showMe.md): The owner wants to see something rather than read it. The agent builds a view for that one question (stats, charts, people, a page per person) and log draws it in the chat with the numbers read from log.db. It reads only and sends nothing. A question about one person gets that person; a question about the week gets the week.

The file

Files
SKILL.md

Showing it as a view: generative UI in the chat

This is the log play inbox/showMe. Before anything else, call the log MCP tool getGuide({ useCase: "inbox/showMe" }): it returns this page and makes it the session's play, so every log tool's answer says which step you are on and a send with no play loaded is refused (rules 157, 158). Then follow the steps in order.

Draft every message from that person's own thread and matt's real replies (voice with no text, search), never from memory, and never send two that read alike (sameAsSent).

The owner wants to see something rather than read it. The agent builds a view for that one question (stats, charts, people, a page per person) and log draws it in the chat with the numbers read from log.db. It reads only and sends nothing. A question about one person gets that person; a question about the week gets the week.

How the owner asks for it: "show me who has been messaging me", "make a report of my DMs", "chart of who messages me most", "a dashboard of my inbox", "tell me about this person", "show me the people who messaged me this week", "visualise my DMs", "graph my messages"

Run it from a fresh context

Say: "go on docs/useCases/inbox/showMe.md" and what to show. The agent reads this page and the ui guide (getGuide({ topic: "ui" })) and runs the steps.

The steps

#StepToolDone when
1Bring the inbox up to date if it may be behind: the view reads log.db, which holds only what has been read (rule 160)twitterDms { account }the read is newer than the question
2Draw it: one call with the data, and code when the question needs its own layout. data: { dms: { source: "twitterDmsBySender", since: "30d" } } alone draws the starter view; add code, a React view against the kit, for anything elsereport { title, subtitle?, data, code? }the answer says the view is shown, or carries it as text
3Fix and call again if the view is refused: the refusal names the line or fieldreportthe view is shown
4Say what it means in a sentence or two, not the numbers againthe replythe owner has the point of it
5Act on what the person clicks: an ask from the view (a follow-up, "draft a reply") arrives as their next message, and a point arrives as contextthe tools that ask needsthe ask is done through its own gates

What to know

  • Where it draws. Claude, ChatGPT and Cursor draw the view inline, because they are MCP Apps hosts. Claude Code, Codex and the rest get the same view as text, which is shown as it is.
  • The view never acts. A button in it can only put a question into the chat or tell the agent what the person is pointing at. Every send still goes through its own tool, its dry run and confirmLive (rule 8).
  • One call is the norm. The data comes in the same call, so the view needs no fetching. useTool inside a view is for data that changes while it is open, and it only calls read tools.
  • The starter view, and when to write your own. With only data, the starter view shows the 30 days in four numbers (messages, the top 10's share, new this week, the busiest day), then the daily line, who messages most, and a table of people that opens a page for each. Write code when the question is narrower than that.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.