Skip to content
Back to search

self-learning-skills

MarketplaceWorks with Claude CodeGitHub

A self-improving skill for AI coding agents (Claude Code, Cursor, AGENTS.md): recognize a hard-won golden path in a session and harvest it into a reusable skill/rule for next time.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/kulaxyz--self-learning-skills/badge.svg)](/listings/kulaxyz--self-learning-skills)

HTML

<a href="/listings/kulaxyz--self-learning-skills"><img src="/api/listings/kulaxyz--self-learning-skills/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

`npx` — recommended (works with 70+ agents)

Uses the community [`skills`](https://github.com/vercel-labs/skills) CLI, which installs into whatever agents it detects — Claude Code, Cursor, Codex, Cline, OpenCode, and more:

npx skills add kulaxyz/self-learning-skills                 # this project (auto-detects agents)
npx skills add kulaxyz/self-learning-skills -g              # global — all your projects
npx skills add kulaxyz/self-learning-skills -a claude-code  # a specific agent

Try it once without installing:

npx skills use kulaxyz/self-learning-skills --skill self-learning | claude

Claude Code plugin

/plugin marketplace add kulaxyz/self-learning-skills
/plugin install self-learning@self-learning-skills

Manual

<details> <summary>Copy the files into place yourself</summary>

git clone https://github.com/kulaxyz/self-learning-skills

# Claude Code — global (or into a project's .claude/skills/ to share via git)
cp -R self-learning-skills/skills/self-learning ~/.claude/skills/

# Cursor — auto-loads .cursor/rules/ (harvested rules land in .cursor/rules/learned/)
mkdir -p .cursor/rules
cp self-learning-skills/.cursor/rules/self-learning.mdc .cursor/rules/

# Any AGENTS.md agent (Codex, Zed, Aider, Gemini CLI, …)
curl https://raw.githubusercontent.com/kulaxyz/self-learning-skills/main/AGENTS.md >> AGENTS.md

</details>

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 30 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "self-learning-skills",
  3. "id": "self-learning-skills",
  4. "owner": {
  5. "name": "kulaxyz"
  6. },
  7. "metadata": {
  8. "description": "A self-improving Claude Code skill: recognize hard-won \"golden paths\" in a session and harvest them into reusable skills for next time.",
  9. "version": "1.0.0"
  10. },
  11. "plugins": [
  12. {
  13. "name": "self-learning",
  14. "source": "./",
  15. "description": "Capture a hard-won golden path from the current session as a reusable Agent Skill, so future sessions start already knowing it.",
  16. "version": "1.0.0",
  17. "author": {
  18. "name": "kulaxyz"
  19. },
  20. "keywords": [
  21. "self-learning",
  22. "skills",
  23. "memory",
  24. "meta-skill",
  25. "workflow"
  26. ],
  27. "category": "workflow"
  28. }
  29. ]
  30. }
README.md · README · 145 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # self-learning-skills
  2. **A self-improving skill for AI coding agents.** Works with Claude Code, Cursor,
  3. and any agent that reads an `AGENTS.md` / standing-instructions file.
  4. Every session you do hard debugging or rediscover the same thing — *how do I
  5. reach the prod DB? where do the creds live? what's the deploy command? how do I
  6. verify this live?* — and that hard-won knowledge evaporates when the session
  7. ends. The next session starts from zero and re-learns it.
  8. **self-learning** fixes that. It teaches your agent to recognize the moment it
  9. has just earned a reusable **golden path** and persist it where the tool will
  10. auto-load it next time — so the next session starts already knowing the route
  11. instead of rediscovering it.
  12. It's a *meta-skill*: it doesn't do the work, it captures **how** the work got
  13. done — including the **failures**, since skipping a known dead-end next session
  14. is often worth more than the win itself.
  15. ## The loop (same everywhere)
  16. 1. **Recognize the moment** — a task that only worked after several tries, a
  17. non-obvious command, a project fact you didn't know up front, an operational
  18. workflow likely to recur, or you simply saying *"remember this"*.
  19. 2. **Capture it, no prompt needed** — it acts on the cue immediately, picks the
  20. scope/name itself, and tells you afterward. The *procedure* is captured (not
  21. a one-off answer), plus a "what didn't work" note.
  22. 3. **Reuse** — next session the entry loads automatically, by skill/rule
  23. description or because the instructions file is always read.
  24. What differs per tool is only *where* knowledge is persisted and *how* it's
  25. auto-loaded:
  26. | Tool | Persists golden paths to | Auto-loads via |
  27. |---|---|---|
  28. | Claude Code, Codex, Agent Skills clients | a new `skills/<name>/SKILL.md` | skill description matching |
  29. | Cursor | a new `.cursor/rules/learned/<name>.mdc` | rule description / globs |
  30. | Zed, Aider, Gemini CLI, … | `AGENTS.md` (or project notes/memory) | always-read instructions |
  31. ## Install
  32. ### `npx` — recommended (works with 70+ agents)
  33. Uses the community [`skills`](https://github.com/vercel-labs/skills) CLI, which
  34. installs into whatever agents it detects — Claude Code, Cursor, Codex, Cline,
  35. OpenCode, and more:
  36. ```bash
  37. npx skills add kulaxyz/self-learning-skills # this project (auto-detects agents)
  38. npx skills add kulaxyz/self-learning-skills -g # global — all your projects
  39. npx skills add kulaxyz/self-learning-skills -a claude-code # a specific agent
  40. ```
  41. Try it once without installing:
  42. ```bash
  43. npx skills use kulaxyz/self-learning-skills --skill self-learning | claude
  44. ```
  45. ### Claude Code plugin
  46. ```
  47. /plugin marketplace add kulaxyz/self-learning-skills
  48. /plugin install self-learning@self-learning-skills
  49. ```
  50. ### Manual
  51. <details>
  52. <summary>Copy the files into place yourself</summary>
  53. ```bash
  54. git clone https://github.com/kulaxyz/self-learning-skills
  55. # Claude Code — global (or into a project's .claude/skills/ to share via git)
  56. cp -R self-learning-skills/skills/self-learning ~/.claude/skills/
  57. # Cursor — auto-loads .cursor/rules/ (harvested rules land in .cursor/rules/learned/)
  58. mkdir -p .cursor/rules
  59. cp self-learning-skills/.cursor/rules/self-learning.mdc .cursor/rules/
  60. # Any AGENTS.md agent (Codex, Zed, Aider, Gemini CLI, …)
  61. curl https://raw.githubusercontent.com/kulaxyz/self-learning-skills/main/AGENTS.md >> AGENTS.md
  62. ```
  63. </details>
  64. ## Triage: skill, memory, or skip?
  65. It won't bloat your config with one-liners. Each lesson is routed:
  66. | Lesson | Where it goes |
  67. |---|---|
  68. | A multi-step, reusable **procedure/workflow** | a new skill / rule |
  69. | A single **fact or one-line correction** | lightweight notes/memory (e.g. a `MEMORY.md`) |
  70. | A genuine **one-off** | skipped |
  71. ## Promotion rule (don't enshrine guesses)
  72. Triage decides *granularity*; the promotion rule decides *confidence*. A skill is
  73. authoritative — the next session trusts it without re-deriving it — so a session
  74. is only promoted to a skill when **all three** hold:
  75. 1. **A passing check** — the path was actually verified (a test passed, a clean
  76. exit, a green build, a reproduced repro). "Seemed to work" doesn't count.
  77. 2. **A named failure pattern** — the failure it avoids or diagnoses, named.
  78. 3. **At least one ruled-out dead-end** — a concrete approach tried and eliminated.
  79. Miss any one and it stays a tentative memory note (or is skipped) rather than a
  80. skill. This keeps confident-but-unverified guesses out of the skill set.
  81. _(Promotion rule suggested by community feedback.)_
  82. ## Safety
  83. Harvested skills/rules get committed and shared, so this is built to **never
  84. write secret values** — no passwords, tokens, connection strings, or API keys.
  85. It records only *where* to find a secret (env var name, a client/selector
  86. function, an MCP tool, a secret manager). Reproducing a secret into a shared
  87. file leaks it.
  88. ## Repo layout
  89. ```
  90. self-learning-skills/
  91. ├── AGENTS.md # generic, cross-tool version of the loop
  92. ├── skills.sh.json # registry manifest for `npx skills` / skills.sh
  93. ├── .claude-plugin/
  94. │ └── marketplace.json # Claude Code plugin manifest
  95. ├── skills/
  96. │ └── self-learning/ # Agent Skills standard (Claude Code + clients)
  97. │ ├── SKILL.md # recognize-the-moment + harvest procedure
  98. │ ├── references/
  99. │ │ └── skill-authoring.md # condensed spec the writer loads to author a good skill
  100. │ └── assets/
  101. │ └── SKILL.template.md # fill-in template for harvested skills
  102. └── .cursor/
  103. └── rules/
  104. ├── self-learning.mdc # Cursor adapter (always-applied rule)
  105. └── learned/ # harvested Cursor rules land here
  106. ```
  107. Built on the open [Agent Skills](https://agentskills.io) standard.
  108. ## License
  109. [MIT](LICENSE) © kulaxyz

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.