at-self-eval

Summarizes Git history and work logs into a review-ready self-evaluation draft.

180|Updated Jul 7, 2026
One-click install
npx skills add https://github.com/kairyou/agent-tools --skill at-self-eval-kairyou
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: at-self-eval
Source: https://github.com/kairyou/agent-tools/tree/main/skills/workflow/at-self-eval
Command: npx skills add https://github.com/kairyou/agent-tools --skill at-self-eval-kairyou

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Writing quarterly, semi-annual, or promotion self-evaluations requires manually digging through months of commits and notes, which is time-consuming and error-prone. This Skill turns your actual Git history and optional work logs into a concise, evidence-backed draft. ## Core Features & Use Cases - Git Evidence Collection: Resolves author and date windows from flags or natural language (e.g., quarters, half-years, month ranges), scans local branches and configured upstreams, and deduplicates commits by hash. - Multi-Source Input: Combines commits across multiple configured repositories with pasted daily/weekly logs or automatically recorded AI session logs, keeping log-only work separate until you confirm it. - Structured Output: Groups related commits into business-line outcomes, folds in housekeeping work, and produces a numbered list of at most 8 concrete results in your language. - Use Case: Before a performance review, ask for a summary of your work from January to June; the Skill scans your repositories, groups the commits into outcomes, and produces a draft you only need to review and polish. ## Quick Start Ask the agent to summarize my Git commits and work log into a self-evaluation for the first half of this year.

Frequently Asked Questions about at-self-eval

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a self-evaluation from my Git commit history?▼

Invoke the skill with an optional author and date range, such as a quarter or half-year. It scans local branches and configured upstreams, groups related commits into business-line outcomes, and outputs a numbered draft of at most 8 results for your review.

How to summarize work across multiple Git repositories for a performance review?▼

Configure workProjects in ~/.agent-tools/config.jsonc with repository paths and optional per-project prompts. The skill resolves each path to a Git root, deduplicates them, scans all involved repos, and merges the evidence into one grouped summary.

Can I use natural language date ranges like last quarter or first half of the year?▼

Yes. Phrases like 统计上半年的工作 or 统计1-3月的产出 are parsed into explicit from/to dates. A period without a year means the current year, except months still ahead of today, which resolve to the previous year.

Does the skill include work that never produced commits?▼

Yes, through optional daily or weekly logs pasted in conversation or configured via dailyLog.output and log.output in the config file. Log-only items are kept separate and included only after you confirm them, using only dates stated in the log.

Why does the skill report no activity for my commits?▼

The author filter matches the exact Git user.name per repository, not aliases or remote logins, and merge commits are always excluded. If nothing matches, the skill reports the author names actually present in the window so you can supply the correct one.

Does it write the self-evaluation to a file automatically?▼

No. The output is a draft printed in the conversation with a reminder that it requires your review. It is only written to a file if you explicitly ask for that.