Skip to content
SkillModel-led2,502 tokens in use

upskill

Reads your past agent sessions and a project's own history, proposes the skills, hooks and automations that would have saved you the most time, then builds your picks.

12,499 stars on the repository MIT Licensejakubkrehel/skills · skills/upskill
npx skills add jakubkrehel/skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing upskill costs

38tokens idle
2,502tokens in use
3,438tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter8/15
  • Structure19/25
  • Examples0/15
  • Output10/15
  • Size15/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity96
  • Maintenance52
  • Community48
  • Maturity24
  • Momentum40

upskill reads every prompt you typed to coding agents on one project, together with the project’s commits, pull requests and CI runs, and counts where your time and patience went. It ranks proposals by what each pattern cost you to recover from, and backs each with a count and dated quotes.

It chooses the lightest mechanism that would hold, from a failing check up to a skill or scheduled run, drops what is already covered, and then builds the ones you pick. You invoke it yourself by command, and a run is meant to take a while.

What You Get

  • A report that opens with findings, each with a count of its own and dated quotes
  • Numbered proposals ordered by cost, each marked small, medium or large, with some marked confirm first
  • A short list of out-of-the-box ideas labelled as guesses, plus what the run would not build and why
  • Built tools: skills, hooks, scripts or agent-doc rules placed by the project’s own conventions
  • A scorecard in later runs showing whether each earlier tool reduced the pattern it targeted

Common Use Cases

  • Finding the corrections you repeat to an agent and turning them into a hook or a rule
  • Replacing a step-by-step procedure you dictate weekly with a skill
  • Spotting CI failures and fix-up commits that a pre-commit check would catch
  • Auditing which installed skills never get used

What Makes It Unique

Proposals rest on counts from the full history and the project record, and the agent re-counts anything a reader reported. A ladder of mechanisms from a failing check to a subagent makes it pick the lightest fix, and a rejection pass asks whether the pattern was the person exploring and not the agent failing. It also keeps your history local: working files go in a git-excluded folder with secrets masked.

Under The Hood

The skill is written instructions plus two reference files, one with the places each agent harness stores sessions and one with the brief given to readers. The agent reads local session logs under your home directory for Claude Code, Codex, Cursor and opencode, and reads the project’s git log, pull requests and CI runs, so it may use the git and GitHub command-line tools. It writes working files into a .upskill folder at the project root and adds that folder to the git exclude file. It masks keys and email addresses in those files and sends no history to outside services, though subagents of the same session read them. After you choose, it writes new skills, hooks, scripts or agent-doc edits into your project and runs hooks and scripts to test them.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers