skill CX Codex

self-improving-agent · pskoett — Claude Code skill: an agent error log

Q: How do I install self-improving-agent in Claude?

Run `npx skills add pskoett/self-improving-agent --skill self-improving-agent -a claude-code`. The skill lands at `~/.claude/skills/self-improving-agent/SKILL.md`.

This skill keeps a markdown log of agent errors, corrections and findings so useful fixes survive between sessions.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓

npx skills add pskoett/self-improving-agent --skill self-improving-agent -a claude-code

Open on GitHub

What self-improving-agent does

Everything worth keeping goes into three files under .learnings/: LEARNINGS.md for corrections, knowledge gaps and good practices, ERRORS.md for failed commands and integration errors, FEATURE_REQUESTS.md for capabilities the user asked for. On first use the directory and files are created; existing ones are never overwritten. Before a big task the agent reads back what it saved and reuses whatever fits.

Entries that turn out to be broadly useful get promoted into workspace files: behavioural patterns into SOUL.md, workflow tweaks into AGENTS.md, tool gotchas into TOOLS.md. Recurring cases carry a Pattern-Key so duplicates don't pile up, and a repeat only counts when it's a genuinely new occurrence. Revalidation is tied to what could change — the source, the decision, a dependency, the task — not to how old a note is.

Secrets, tokens and environment variables stay out of the log; short summaries or redacted excerpts are preferred. Automatic reminders and end-of-session error sweeps come from an optional hook, and it only runs where .learnings/ already exists. This build targets OpenClaw; for other agents the author points to the original multi-agent version in a separate repo.

What it helps with

log a failed command so it doesn't bite twice
save a user correction as a working rule
record a missing capability and revisit it later
pull a good approach from an earlier session into the current task
move a recurring tool gotcha into workspace notes

Example prompts for the agent

log this failure to .learnings/ERRORS.md
the user corrected my assumption — record it as a correction
check .learnings/LEARNINGS.md before you start the refactor

How to install self-improving-agent

1Skills CLI (Vercel)

npx skills add pskoett/self-improving-agent --skill self-improving-agent -a claude-code

2Manually: put the files in place

~/.claude/skills/self-improving-agent/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/self-improving-agent/

1Skills CLI (Vercel)

npx skills add pskoett/self-improving-agent --skill self-improving-agent -a codex

2Manually: put the files in place

.agents/skills/self-improving-agent/SKILL.md

For all projects use ~/.agents/skills/

1Skills CLI (Vercel)

npx skills add pskoett/self-improving-agent --skill self-improving-agent -a cursor

2Manually: put the files in place

.cursor/skills/self-improving-agent/SKILL.md

Cursor also reads .agents/skills/ and .claude/skills/

1GitHub CLI

gh skill install pskoett/self-improving-agent self-improving-agent

2Manually: put the files in place

.github/skills/self-improving-agent/SKILL.md

1Skills CLI (Vercel)

npx skills add pskoett/self-improving-agent --skill self-improving-agent -a gemini-cli

2Platform CLI

gemini skills install https://github.com/pskoett/self-improving-agent --path self-improving-agent

Manually: put the files in place

.devin/skills/self-improving-agent/SKILL.md

The legacy .windsurf/skills/ path still works

Skills CLI (Vercel)

npx skills add pskoett/self-improving-agent --skill self-improving-agent -a opencode

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about self-improving-agent

How do I install self-improving-agent in Claude?

Run npx skills add pskoett/self-improving-agent --skill self-improving-agent -a claude-code. The skill lands at ~/.claude/skills/self-improving-agent/SKILL.md.

What does self-improving-agent do?

It keeps errors, corrections and knowledge gaps in markdown files under .learnings/, then promotes the important ones into workspace files such as AGENTS.md and TOOLS.md.

Does it need any companion service or plugin?

No. Capture, review, reuse and promotion all live inside the skill. The hook is optional: it reminds you and sweeps errors at session end, but it doesn't enforce the workflow.

self-improving-agent by other authors comes first.

self-improving-agent Alireza Rezvani Curates Claude Code auto-memory from MEMORY.md, promotes proven patterns to CLAUDE.md, and extracts reusable skills from recurring solutions. 1,078installs self-improving-agent zhaono1 Turns workflow failures, corrections, and workarounds into auditable rules for agents. Prevents blind instruction rewrites by requiring executable proof first. CX Codex +1 35,100installs self-learning kulaxyz A meta-skill that captures a working route from the current session as a reusable skill, so the next session starts already knowing it. CX Codex +1 850installs blueprint affaan-m Turns a one-line objective into a step-by-step construction plan: PR-sized steps, a dependency graph, parallel branches and an adversarial review pass. CX Codex +1 4,600installs planning-with-files zhaono1 Keeps the plan, research notes and final output in markdown files instead of context. Useful for multi-step tasks and research work. CX Codex +1 1,200installs desk-open official github Opens a new desk in the workshop: a folder, a first journal entry and a signals directory, so the next Copilot session finds the trail. 23installs self-hosted-funnel-launch autonnel This skill takes a self-hosted funnel from an empty install to published pages, checkout, an upsell and a thank-you step. Use it when an agent must deploy and edit the funnel over MCP. CX Codex +1 103,000installs convex-self-heal convex The skill takes a production error through triage, root cause and repair, certifies the fix on a preview, then opens a PR. A human merges. CX Codex +1 35,800installs