agentleFS
matthewdonsemail-labmatthewdonsemail-lab/log/skills/console/SKILL.md

console

Skill63 starsChanged yesterdayMITView on GitHub

Read this machine's local console, the same data in the terminal, as JSON per screen, or in the web console served on loopback with no sign-in. Use when someone asks "is log set up here", "is the MCP running", "which accounts are signed in", "how many calls today", "show the latest events", "what MCP config do I paste into Claude Code / Codex / opencode", or "open the console".

  • Reads credentials

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every screen takes --json.

    line 23
  2. Very specific:

    Read the screen that answers the question: Use the MCP tool (consoleStart, consoleEvents …).

    line 27

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

console

log console is the local console. Everything it shows is read on this machine: log.db, ~/.logCli/localAccounts.json, each account's governor and the MCP config. There is no sign-in, no team and no billing locally; a team needs log login.

Screens

One screen per web console nav item, in the same order, each backed by one loader in src/console/<domain>/:

ScreenTerminal and MCPWeb (log console --web)Loader
StartconsoleStart / log console start/console/startsrc/console/start
ConnectionsconsoleConnections / log console connections/console/connectionssrc/console/connections
UsageconsoleUsage { days: 1, 7 or 30 }/console/usagesrc/console/usage
EventsconsoleEvents { kind, status, page, pageSize }/console/eventssrc/console/events
AI clientsconsoleAiClients { client }/console/mcpsrc/console/aiClients
SettingsconsoleSettings/console/settingssrc/console/settings

Every screen takes --json; the MCP tools return that JSON. Teams and Subscription say "local: no team or billing" and point to log login.

The loop

  1. Read the screen that answers the question. Use the MCP tool (consoleStart, consoleEvents …). It reads only; it never signs in, sends or resets anything.
  2. Act through the command it names. Connections lists accounts and their session state; checking or signing one in is log <platform> session check or log <platform> cookies save, which the interactive console runs for you through the same code as log launch.
  3. The terminal console is log console (or log launch, Console). It needs a terminal, so an agent reads the screens instead.
  4. The web console is log console --web [--port <n>] [--noOpen]. It serves the built site (site/apps/web/dist) on 127.0.0.1 (port 7338 by default), answers /api/console/* from the same loaders, and opens /console/start. The page knows it is local because the server injects window.logConsoleMode = "local" into the page, and /api/console/mode answers { mode: "local" }; it then skips the sign-in and team redirects. If the site isn't built it prints the build command: pnpm --dir site install && pnpm --dir site --filter web build. For the Vite dev server, run log console --web first and start Vite with LOG_CONSOLE_API=http://127.0.0.1:7338: it marks the page local and proxies /api/console there.
  5. The hosted console (a signed-in team, e.g. log.listeningkit.com) shows the same screens from the team's paired workspace: the control service relays /api/v1/teams/<team>/console/* to log runtime connect, which answers with the same src/console/operations (connectionsList, credentialsSave, credentialsRead, accountCheck, accountSignIn, usage, events). The site's connectionsData, usageData and eventsData pick local or hosted, so no screen cares which.
  6. What the web Connections screen changes. Manage keys checks a key with its provider, then saves it to ~/.logCli/.env on the workspace (AdsPower: its Local API /status; an empty AdsPower key keeps the held one). Check status runs <platform>SessionCheck, one at a time. Sign in with pasted cookies runs the platform's own log <platform> cookies save (X, LinkedIn, Slack); a Facebook username and password are saved where log facebook connect reads them and the browser check finishes the sign-in. Other platforms refuse rather than pretend.
  7. The Facebook browser check is the Connections row's Browser check (and what a Facebook password sign-in opens). It drives the account's one browser through the console operations browserState, browserOpen (inspectStart, detached), browserFrame and browserInput (the live viewer and its gated /input), browserAction (back and reload through inspectDo, verify with a snapshot, resolve with inspectDoLive signIn from the saved login, link profile with adsPower sync), browserSaveState (inspectDo saveState, then the account shows as checked) and browserClose (inspectDo stop, only for a browser the dialog itself opened). Local: the browser is also open in the window on the screen, and the console shows its frames. Remote: voxl streams it when the runtime has one, frames otherwise; the stream descriptor goes only to the dialog and is journalled as <redacted>, like frames.
  8. Hosted widgets. The control service's /mcp gives a client that advertises io.modelcontextprotocol/ui (remembered in a signed Mcp-Session-Id, nothing stored) the runtime's ui:// resources (uiResourcesList, uiResourceRead), browserView with its _meta.ui, and the app-only tools (browserViewStream, browserViewFrame, browserViewInput, browserViewDone, visibility: ["app"]); other clients see none of them and are refused. App-tool answers are relayed, never recorded.

Testing the hosted path

Pair a second runtime that cannot touch the real one: its own LOG_RUNTIME_DIR and its own log.db (LOG_DB or LOG_INBOX_STATE in the same temporary folder). log runtime connect refuses a LOG_RUNTIME_DIR without one. Point it at a local control service on a temporary SQLite DATABASE_URL, and never at the hosted one. Account checks and sign-ins are not run against real accounts while testing.

Hard boundaries

  • Loopback only. The server binds 127.0.0.1 and refuses any request whose Host, remote address or Origin isn't this machine, or which is cross-site, the same guard as the dev server's /api/console/events. Changes are same-origin JSON posts with the x-log-console: 1 header.
  • Keys never leave the workspace: answers carry only hasValue and a fixed mask, credentialsRead answers AdsPower's countries and URL only, secret arguments are journalled as <redacted>, and the control service stores none of it (rule 172).
  • Governor resets stay with governorReset and the owner's word (rule 168); sends stay with their own tools and gates.
  • Parity is checked: bun run check:consoleParity fails when a web nav item has no terminal screen, or the two don't read the same src/console/<domain> loader.

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.