Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add berekvolgyipeter/dotclaude --skill learngit clone --depth 1 https://github.com/berekvolgyipeter/dotclaudeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/berekvolgyipeter/dotclaude/learn)<a href="https://agentmods.dev/skills/berekvolgyipeter/dotclaude/learn"><img src="https://agentmods.dev/badge/skills/berekvolgyipeter/dotclaude/learn/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/berekvolgyipeter/dotclaude/learn"><img src="https://agentmods.dev/badge/skills/berekvolgyipeter/dotclaude/learn.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00119 | $0.02628 |
| Opus 5.5 | $0.00048 | $0.01051 |
| Sonnet 5.5 | $0.00024 | $0.00526 |
| Haiku 4.5 | $0.00012 | $0.00263 |
Grade A, and why
learn scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Triage first
Before diving into analysis, check if there's anything to learn. Most conversations have zero learnings; a few have one or two. Fabricating learnings from uneventful interactions creates noise that drowns out real insights — say so and stop if the conversation was routine.
A single offhand correction in an otherwise smooth conversation may not warrant a rule. Focus on patterns that would recur across future sessions.
Analysis Process
Learnings live at two levels:
- User-level (
~/.claude/) — shared across all projects (rules, skills, commands, agents, output styles, templates) - Project-level (
.claude/) — specific to the current project (CLAUDE.md, rules, skills, commands, agents)
Step 1: Scan for Learnings
Examine the conversation for learning signals and identify rule gaps in a single pass:
- User corrections: "Actually, you should...", "No, the correct way is...", "Don't do X, do Y instead" — includes cases where the user had to repeat themselves or rephrase
- Agent mistakes: Repeated errors, wrong assumptions, misunderstandings, or violations of existing rules (needs new rule or emphasis/clarification of existing one)
- Emergent patterns: Something worked well that should be documented, or a workflow crystallized that future sessions would benefit from. Positive learnings go to the same target files as fixes — add them as recommended patterns rather than corrective rules.
- Tool invocations: Instances of
Skilltool calls (skills/commands) andTasktool calls (agents) — these feed into Step 2
Step 2: Evaluate Used Skills, Commands, and Agents
Scan the conversation for any invocations of custom tools:
Skills — Skill tool calls where the skill parameter matches a name in ~/.claude/skills/ or .claude/skills/
Commands — slash command invocations like /fix-review, /to-plan, etc., which expand via the Skill tool
Agents — Task tool calls where subagent_type matches a name in ~/.claude/agents/ or .claude/agents/
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 190 lines · 119 tokens per session scan A 651ba3394d6a
learn is a skill published in the GitHub repository berekvolgyipeter/dotclaude (2 stars, last pushed 3d ago), licensed MIT. It adds 119 tokens to every session and 2,628 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
accint-solve
Route agent work through AccInt's MCP memory loop: retrieve prior outcomes, resolve frames, and close commitments with evidence.
brainstorming
Interactive spec generation — turn ideas into concrete specs with R-numbered requirements and testable acceptance criteria.
executing
Autonomous task implementation workflow — implement, test, review, commit for each task in the frontier.
backpropagation
Trace runtime bugs back to spec gaps — identify missing acceptance criteria, update specs, generate regression tests, and detect patterns.
collaborating
Hackathon-native multiplayer collaboration mode -- brain-dump together, claim tasks across machines, flag decisions async. Drives the subcommands of /forge:collaborate.
caveman-internal
Internal token optimization for Forge agent artifacts (handoff notes, artifact summaries, review notes). NOT exposed as a user-facing /caveman command.