Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

Run History Skill Upgrader

Turns real run evidence, recurring failures and user feedback into a concrete upgrade plan for an existing skill, applying changes only after explicit approval.

UtilitiesIntermediate★ 749⑂ 61AI score 8/10Last updated: Aug 26, 2026

What it does

  • Gathers real evidence — run logs, screenshots, diffs, test results, official docs, platform changes, and user corrections — and grades it as strong or weak.
  • Classifies signals (process gap, validation gap, source drift, platform drift, preference, incident, routing gap, content bloat) and runs a generalization gate so one-off incidents don't become permanent rules.
  • Picks a net-improvement shape: no_change, maintenance_note_only, prune_or_consolidate, local_refactor, cross_reference_refactor, major_refactor, or deprecate_or_replace_source.
  • Starts in plan-only mode by default and stops; "go ahead" or "don't ask me" do not count as approval.
  • After approval, edits only in-scope files, then runs the bundled validator plus py_compile, json.tool, and route/boundary regression, and reports what changed and what risk remains.

Who it's for

  • Authors maintaining a growing library of agent skills (Claude Code, Codex, OpenCode, and similar hosts).
  • Teams sharing skills who need an approval gate before any file is rewritten.
  • Anyone whose SKILL.md keeps accumulating reminders instead of being pruned.

Example uses

  1. "Our deploy skill failed the same validation three times this week — here are the logs, plan a fix."
  2. "The vendor API moved to v2 but our skill still documents v1" → classified as source drift, with stale guidance merged or deleted.
  3. "This skill is bloated" → machine-checkable facts get moved into scripts/tests and the prose is consolidated.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/dongshuyan/compass-skills/HEAD/skills/run-history-skill-upgrader/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/run-history-skill-upgrader folder from the GitHub repo dongshuyan/compass-skills into my ~/.claude/skills/run-history-skill-upgrader/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/dongshuyan/compass-skills.git && mkdir -p ~/.claude/skills && cp -r compass-skills/skills/run-history-skill-upgrader ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal.
  2. Clone the repository: git clone https://github.com/dongshuyan/compass-skills.git
  3. Create the skills folder: mkdir -p ~/.claude/skills
  4. Copy the skill: cp -r compass-skills/skills/run-history-skill-upgrader ~/.claude/skills/
  5. Verify Python is available: python3 --version (needed for the bundled validator).
  6. Restart Claude Code and ask it to "improve skill X based on recent runs and failures".
  7. Review the plan it returns, then approve explicitly (e.g. "apply the plan above") before any files are modified.
View source on GitHub ↗License: MIT