# Mk Observe

> Collect, score and browse agent activity across Claude Code, opencode and Kiro CLI — four honest dimensions (acceptance, success, autonomy, reliability) with formulas and provenance, token usage per model (raw counts, no price guessing), kit adoption (which commands/skills/agents/hooks a team really uses, per project), workflow and outcome facts (prompt bands, steering, compactions, commits, tests, PRs) and rule-based insights for team leads; CLI tables and a local web UI. Use when asked how the agent/harness is performing, how many tokens a week/model/project burned, whether the team uses the kit, which hooks fire or fail, or to compare harnesses/models/agents on the same project.

- Skill: `phuc-nt/mk-observe-2` (Agent Skill)
- Install (CLI): `npx skillmds@latest add phuc-nt/mk-observe-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/phuc-nt/mk-observe-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: phuc-nt (https://skillmd.com/u/phuc-nt)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/phuc-nt/mk-observe-2

---

# mk-observe

This is a wrapper for the kit skill in `.claude/skills/observe/`.

1. Read `.claude/skills/observe/SKILL.md` with the file reading tool.
2. Follow its instructions exactly for the request below. Relative paths it mentions (`references/`, `scripts/`, `bin/`, `assets/`) resolve from `.claude/skills/observe/`.
3. Where it says to invoke another `/mk-*` skill, load that skill the same way.

Argument hint: `[stats|runs|run <id>|serve|export|daemon install|status|uninstall|store status|prune] [--days N] [--cwd path] [--all]`

Request / arguments: $ARGUMENTS