Skip to content
mcp/skillhub

Agent Skill

Compound Learnings skill

compound-learningsparcadei/Continuous-Claude-v3.claude/skills/compound-learnings/SKILL.md

Transform session learnings into permanent capabilities (skills, rules, agents). Use when asked to "improve setup", "learn from sessions", "compound learnings", or "what patterns should become skills".

3.9kMIT

always on

45 tokname + description

once used

1,597 tokSKILL.md body

bundled files

0

repo stars

3.9kGitHub

Install the Compound Learnings skill

Skills CLI (Claude Code, Codex, Cursor and others)

npx skills add https://github.com/parcadei/Continuous-Claude-v3 --skill compound-learnings

Claude Code, by hand (all projects)

git clone --depth 1 https://github.com/parcadei/Continuous-Claude-v3.git /tmp/Continuous-Claude-v3
mkdir -p ~/.claude/skills
cp -R /tmp/Continuous-Claude-v3/.claude/skills/compound-learnings ~/.claude/skills/compound-learnings

For one project only, copy the folder into .claude/skills/ in that project instead. The agent picks it up when a task matches its description.

What it costs your context

Agent Skills load in three steps. Tokens counted with the o200k tokenizer; your model's count will differ a little.

WhenWhat loadsCost
Every sessionName and description45 tok
When the agent uses itThe SKILL.md instructions1,597 tok
Only if openedNo bundled files—

Until it is used, compound-learnings costs about 30× less than the median MCP server we measured (1,363 tokens on every turn). Installing all 150 skills from parcadei/Continuous-Claude-v3 would load 2,175 tokens of descriptions. Skills vs MCP study

Checks

Our checks found nothing to flag: it follows the Agent Skills specification's naming and size guidelines and doesn't pipe downloads into a shell.

Pre-approved tools (allowed-tools): Read, Glob, Grep, Write, Edit, Bash, AskUserQuestion. The agent can use these without asking while the skill is active.

Checked 2026-10-22 · method

What's inside

The headings of its SKILL.md. Read the full file on GitHub.

  • Compound Learnings
  • When to Use
  • Process
  • Step 1: Gather Learnings
  • Step 2: Extract Patterns (Structured)
  • Step 2b: Consolidate Similar Patterns
  • Step 3: Detect Meta-Patterns
  • Step 4: Categorize (Decision Tree)
  • Step 5: Apply Signal Thresholds
  • Step 6: Propose Artifacts
  • Step 7: Create Approved Artifacts
  • Step 8: Summary Report
  • Quality Checks
  • Files Reference

More skills from parcadei/Continuous-Claude-v3

All skills in this repo
  1. agentica-claude-proxyparcadei/Continuous-Claude-v3

    Guide for integrating Agentica SDK with Claude Code CLI proxy

    3.9k19 tok idle · 1,005 in use
  2. agentica-infrastructureparcadei/Continuous-Claude-v3

    Reference guide for Agentica multi-agent infrastructure APIs

    3.9k14 tok idle · 686 in use
  3. agentica-promptsparcadei/Continuous-Claude-v3

    Write reliable prompts for Agentica/REPL agents that avoid LLM instruction ambiguity

    3.9k21 tok idle · 1,701 in use
  4. agentica-sdkparcadei/Continuous-Claude-v3

    Build Python agents with Agentica SDK - @agentic decorator, spawn(), persistence, MCP integration

    3.9k23 tok idle · 2,798 in use
  5. agentica-serverparcadei/Continuous-Claude-v3

    Agentica server + Claude proxy setup - architecture, startup sequence, debugging

    3.9k18 tok idle · 1,062 in use
  6. analytic-functionsparcadei/Continuous-Claude-v3

    Problem-solving strategies for analytic functions in complex analysis

    3.9k12 tok idle · 952 in use
  7. ast-grep-findparcadei/Continuous-Claude-v3

    AST-based code search and refactoring via ast-grep MCP

    3.9k17 tok idle · 673 in use
  8. banach-spacesparcadei/Continuous-Claude-v3

    Problem-solving strategies for banach spaces in functional analysis

    3.9k15 tok idle · 905 in use