# Observe

> Show whether the agent followed the kit's rules in recent sessions: pages edited and measured, Stop hook blocks, findings left open. Reads local hook events only.

- Skill: `tuyv/observe` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx get-skill add tuyv/observe`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tuyv/observe/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: tuyv (https://skillmd.com/u/tuyv)
- Updated: 2026-11-14
- Page: https://skillmd.com/skills/tuyv/observe

---

## Requirements

This skill needs the full UX/UI Agent Skills kit: the gates, the references and
the design-system library. Install it with `npx ux-ui-agent-skills@3.1.0 init`
or as the Claude Code plugin. `npx skills@1.7.6 add` copies the skill folder and
nothing else, and this skill has no guidance-only mode: its whole job is to
read what the kit's hooks recorded, so without the kit there is nothing for it to do.

> **Step 0: is the kit here?** Check once:
> `ls ${CLAUDE_SKILL_DIR}/../../../scripts/observe.mjs >/dev/null 2>&1 && echo KIT_OK || echo KIT_MISSING`
> On `KIT_MISSING`, say plainly that only the skill folder was installed, name
> the install command above, and stop. Do not guess the contents of a file you
> could not open, and do not report a result you did not produce.

Report what the hooks recorded - never a reconstruction from memory.

1. Run `node ${CLAUDE_SKILL_DIR}/../../../scripts/observe.mjs --project "$CLAUDE_PROJECT_DIR"`
   (or without `--project` when the user asks about every project, and with
   `--days N` when they name a window).
2. Quote the summary lines it prints. If it says no session edited UI, say that
   and stop: there is nothing to judge.
3. For each session marked `BROKE A RULE`, run it again with `--session <id>` and
   say which rule failed and the timeline line that shows it.
4. If the user wants the page, run it with `--html <path>` and give the path.

What it measures, and what it does not: three counted rules per session -
every edited page had a passing gate before the end, no fast-gate finding was
left on a file's latest version, and the agent never finished while the Stop
hook still had pages open. It says nothing about design quality, and an
`unfinished` session is not counted for or against.

