←Leaderboard/Code Quality/paperthin
Last commit on November 9, 2026·Created on June 19, 2026

LilMGenius/paperthin

“Implement recursive self-improvement for coding agents.”
Combined rank
#326
across all skills
In Code Quality
#19
category rank
Stars
1.4k
+1.8% in last 7d
Forks
121
Watchers
5
Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
TL;DR

Paperthin provides a skillbook designed for agentic engineering and recursive self-improvement.

WHO IT'S FOR
AI agent / automation builders
implement recursive self-improvement for agents
Platform / DevEx teams
standardize agent-facing documentation and manifests
Technical leads / architects
design high-fidelity knowledge base routing
Claude Code or Codex users
manage and update agent skill installations
Repository contents

45 skill files

Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.

Use this skill collection

Follow the documented setup, then try a first task.

I’m using

Codex plugin

Installs: The paperthin plugin bundle, documented to update at every session start. · Codex

Before you start
  • Codex CLI available.
  1. Run this command in a terminal to add the marketplace and install the paperthin plugin.

    codex plugin marketplace add LilMGenius/paperthin && codex plugin add paperthin@paperthin
README.md ↗ · Checked Nov 9, 2026

Give it something to do.

Suggested first task

Turn supplied material into a proposed intent

Uses aim

  • Attach or paste material such as files, notes, or conversation history whose intended next step is not yet clear.

Make sure aim is installed in your agent before trying this task.

Use aim to read the material I attached end to end. Propose the most likely intent in one grounded sentence, briefly say what you would do next, and ask me to confirm or adjust it.
skills/depth/aim/SKILL.md ↗ · Checked Nov 9, 2026
Suggested first task

Find one next action for a stalled project

Uses nba

  • Provide access to the project's current plan, priority notes, recent changes, and available QA or workflow evidence.

Make sure nba is installed in your agent before trying this task.

Use nba to read this project's current state and recommend one next best action without executing it. Cite the evidence behind the choice and explain where we are, why this action matters now, and what observable result would mean it is done.
skills/coil/nba/SKILL.md ↗ · Checked Nov 9, 2026
Suggested first task

Check an evaluation for leakage

Uses mandela

  • Attach a validation or evaluation design describing its model, scorer, designer, dataset, and how success is measured.

Make sure mandela is installed in your agent before trying this task.

Use mandela to audit the attached validation design in its default read-only mode. Identify its model, scorer, designer, and dataset; check the eight leakage patterns; and report only patterns that fire, with evidence and an independent-ground-truth fix for each. Do not rewrite the experiment.
skills/depth/mandela/SKILL.md ↗ · Checked Nov 9, 2026