agentsclimarketplace

Activity view

Skill chicongst/kortex/skills/activity-view

Open the Activity Tracker live viewer (KORTEX) that animates what Claude is doing right now — a flow graph of prompts, tools, and the parallel subagent tree, plus tokens/cost and context usage. Use when the user wants to watch, monitor, or visualize Claude Code activity.From its SKILL.md

Install
npx -y skills add chicongst/kortex --skill activity-view

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 8 stars8 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.7 KB, as published. Nobody here has run it

Activity View (KORTEX)

The Activity Tracker plugin streams every tool call, file edit, command, and subagent event from Claude Code to a local viewer that animates it in real time.

Start the viewer

Usually you don't need to — the plugin autostarts the viewer on SessionStart (scripts/ensure-viewer.sh). If it isn't running, start it manually:

node "${CLAUDE_PLUGIN_ROOT}/viewer/server.js"
# custom port must match the plugin's `endpoint` userConfig:
node "${CLAUDE_PLUGIN_ROOT}/viewer/server.js" --port=39000

Then open:

Events flow through command hooks (scripts/emit.sh) to the viewer's /event; to use a different port, set the endpoint userConfig (the autostart and toggle derive the port from it).

What you see (flow graph)

  • PRIMARY core — the session reactor, ringed by a context-window gauge (fills cyan → amber → red as the prompt fills the model's window).
  • Thread nodes — one per prompt; the newest is biggest and reaches farthest out, older ones shrink and pull in. Each turn's tools branch off its thread.
  • Tool nodes — Read/Edit/Bash/… beads; slower steps are bigger, failures turn red.
  • Subagents — the parallel subagent tree, each a violet branch with its own tools.
  • Left SYSTEM panel — sessions switcher, alerts, token usage (input/output/cache), context %, cost drivers, and an event log. Right INSPECTOR — click any node for its per-node token/cost breakdown.
  • SAID popup — the text Claude wrote on the current turn, beside the active thread.

Tokens & cost — zero config

Cost, tokens, model, and context % are read automatically from the session transcript (no setup). OpenTelemetry is optional — a fallback for cost/tokens when a transcript isn't available, plus HTTP-level signal the transcript lacks (API errors, 429/529, retries):

export CLAUDE_CODE_ENABLE_TELEMETRY=1
export OTEL_METRICS_EXPORTER=otlp
export OTEL_EXPORTER_OTLP_PROTOCOL=http/json
export OTEL_EXPORTER_OTLP_ENDPOINT=http://127.0.0.1:39000

Turn it off / on

/activity-tracker off (or on, status), or click the ⏻ power switch in the viewer's top-left. The choice persists across sessions.

Keep looking

Skills are one crate of 610,677. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.