# Univer Worklog Auto

> Use when the user wants to automatically collect Codex/workbuddy/git/GitHub evidence, summarize work, dedupe it, and write or dry-run Univer worklog rows. Supports --dry-run, --confirm, --no-submit, and --period day|week|month.

- Skill: `dream-num/univer-worklog-auto` (Agent Skill)
- Install (CLI): `npx skillmds@latest add dream-num/univer-worklog-auto`
- Raw SKILL.md: https://api.skillmd.com/api/skills/dream-num/univer-worklog-auto/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: dream-num (https://skillmd.com/u/dream-num)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/dream-num/univer-worklog-auto

---


# univer-worklog-auto

Use the core protocol in `../univer-team-standup/SKILL.md`, route intent `auto`.

Support:

- `univer-worklog-auto`
- `univer-worklog-auto --dry-run`
- `univer-worklog-auto --confirm`
- `univer-worklog-auto --no-submit`
- `univer-worklog-auto --period day|week|month`

Follow the core Auto workflow exactly: bootstrap Univer dependencies if missing, enforce the mandatory team remote `fYmh0HRyTUO6YECQGFScnA0` on `https://univer.ai/`, collect evidence, summarize, resolve priority through `WorkItems`, dedupe with `去重键`, write allowed rows, verify by reading back, audit, preview, and auto-submit when the Team Auto Submit policy allows it. Auto must re-check the bound unit id before pull, before writing, before commit/sync, and after sync; if the target cannot be proven as `fYmh0HRyTUO6YECQGFScnA0`, stop before syncing. Always follow the core Personal Log Date Field Policy: write `日期` as the local work date, apply `yyyy-mm-dd` display formatting, and verify the date display is not a naked spreadsheet serial number. Always follow the core Progress And Risk Prompting Policy: infer progress, risk, blocker, plan, and next-action fields from evidence when possible, and surface missing items as `需补充` or `待确认` before writing.

During 2026-05-21 through 2026-06-18, apply the core Business Goal Tracking rules when summarizing candidates. If evidence clearly maps to an O1-O4 KR, use goal-aware `关联项` or `WorkItems` hints; if uncertain, leave the candidate unmapped and let reports surface the possible focus drift.