Skip to content
Get packs

HillClimb Improver

Improves a LifeOS skill, prompt or workflow by testing changes on held-out cases and keeping only proven wins

Source Download skill
Repository
danielmiessler/LifeOS
Stars
19,682
Licence
MIT
Files
15 files

What it does

Once installed, the agent can take one named thing in a LifeOS setup (a skill, workflow, eval, prompt, rule or code metric) and improve it through a structured experiment loop. It gathers complaints and failing cases, proposes varied changes, measures each against held-out test cases, keeps only changes the measurements support, and records every attempt in a ledger. Each win is turned into a standing check so the same failure does not return. Explaining it simply: instead of guessing at edits, the agent runs a disciplined trial-and-error process and reports what actually helped.

Great for

  • Developers and power users who maintain LifeOS skills, evals or prompts and want measured improvements
  • Anyone who wants a recurring failure fixed permanently and verified with before-and-after numbers
  • Teams running evaluation suites who want changes gated by held-out scores rather than opinion

Skip it if

  • Filing a one-off complaint or bug report, where the Errata skill is the right tool; it is not needed for simple logging
  • One-pass brainstorming or quick ideas, which add cost without measurement benefit
  • Anyone without the LifeOS instruments (Errata, Assay, ISA) in place, since the skill depends on them and on local run files; results would be unreliable or unavailable

Try asking

  • “Climb this skill: improve the Research skill's output quality and keep only changes that hold on held-out cases”
  • “What hurts most in my LifeOS setup? Survey the hills and pick one to improve”
  • “This keeps failing, fix it for good: turn this miss into a standing check and verify the complaints fell”
  • “Reduce latency of this workflow with a before-and-after measurement”
Step-by-step guide: how to use HillClimb ImproverInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 15 original files, the MIT licence and an attribution note.

    HillClimb.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip HillClimb.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip HillClimb.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Start with a clear target and a stated budget (for example start --budget 10) so the climb has a defined stop
  • Let the survey command rank real hills first, then pick one, rather than choosing a target by hand
  • Review the ledger and the kept diff before landing changes, since scripts run locally and the climb records every experiment

What's inside

  • Intents/Eval.md
  • Intents/Ideate.md
  • Intents/IsaClimb.md
  • Intents/Loop.md
  • Intents/Metric.md
  • Intents/Miss.md
  • Intents/Prompt.md
  • Intents/Skill.md
  • Intents/Survey.md
  • Intents/Verify.md
  • Intents/Workflow.md
  • References/CritiqueClimb.md
  • References/Experiments.md
  • References/Instruments.md
  • SKILL.md

Redistributed unmodified from danielmiessler/LifeOS under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.