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 Osipchuk/agent-skills --skill learning-modegit clone --depth 1 https://github.com/Osipchuk/agent-skillsWrote 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/osipchuk/agent-skills/learning-mode)<a href="https://agentmods.dev/skills/osipchuk/agent-skills/learning-mode"><img src="https://agentmods.dev/badge/skills/osipchuk/agent-skills/learning-mode/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/osipchuk/agent-skills/learning-mode"><img src="https://agentmods.dev/badge/skills/osipchuk/agent-skills/learning-mode.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 | $0.00235 | $0.06648 |
| Opus 5.5 | $0.00094 | $0.02659 |
| Sonnet 5.5 | $0.00047 | $0.01330 |
| Haiku 4.5 | $0.00023 | $0.00665 |
Grade B, and why
learning-mode scanned grade B with 2 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.
Strips warnings and disclaimersmediumAnti-refusal
Omitting safety caveats hides risk from the user and is a common jailbreak preamble.
- Don't gold-plate. If they wrote a one-off script, don't lecture them about dependency injection. Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
results.append(requests.get(url).json()) How it starts
The opening of the file, as written. The whole thing — 405 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Learning Mode
A skill that turns regular development sessions into deliberate practice. The user keeps shipping real work in a real repo, but at well-chosen moments hands off a piece to write themselves while Claude steps back, then reviews. State lives in .claude/learning/ so it survives sessions and stays human-editable. Whether to commit it is the user's call, not a default: unlike a team's scars log or onboarding config, this state is one developer's — goals, a self-rated level, and a dated record of what they got wrong. In a shared repo it collides between developers and exposes more than most people want reviewed; gitignoring .claude/learning/ is the safer default there. Say so once if the repo has no rule yet, then respect whatever they choose.
Mental model
Two layers running together: normal development as usual, plus a coaching overlay where Claude leaves one small atomic gap for the user to fill, with a brief sized to their level. How much Claude writes around the gap depends on the task and the user's level — see "Handoff modes".
The atom is the contract: one function body, one validator, one regex, one small algorithm. Not "an endpoint" or anything plural. Roughly 5–15 minutes for the user's level. Feel like a senior pair-programmer, not a teacher assigning homework.
The handoff is a split of the work, not a pause in it. The user takes the atom; you take the rest of the same feature. Claude Code is turn-based — you cannot type while the user types — so the way this actually works is ordering inside one turn: hand off first, then write your own slice, then yield. Never end your turn on the handoff itself. The user should leave to solve their atom and come back to a repo that moved forward, not to a session that sat waiting for them.
State files
Everything lives under .claude/learning/ in the project root:
.claude/learning/
├── plan.md # Goals, focus areas, engagement mode, level
├── progress.md # The spaced-repetition log (table of topics)
└── active-task.md # Brief for the currently-handed-off task (when one is in flight)
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 405 lines · 235 tokens per session scan B c06297c5ae24
learning-mode is a skill published in the GitHub repository Osipchuk/agent-skills (5 stars, last pushed 2mo ago), licensed MIT. It adds 235 tokens to every session and 6,648 once invoked, about $0.0009 per session on Opus 5.5. A static security scan graded it B with 2 findings (strips warnings and disclaimers, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
socratic-quiz
Use this when the user wants to deeply understand something through guided questioning. Trigger phrases include: "quiz me", "help me understand", "Socratic", "teach me", "walk me through with questions", "test my understanding", or when the user asks for an explanation and would benefit more from guided discovery than…
take-notes
Turn a YouTube video, a web article, or several sources at once into one set of didactic study notes, written as a self-contained HTML page under /take-notes/htmlreports and opened in the browser.
skill-powerups
Interactive tutorial-style lessons for custom skills. Invoke /powerup for a guided walkthrough with examples, exercises, and knowledge checks. Trigger: When user says "powerup", "tutorial", "teach me", or wants to learn a skill interactively.
simplicio-learn
Persist what a run taught you so the next run is cheaper and more correct — mine high-signal lessons from the trajectory, dedup them, and write them back to AGENTS.md / memory so they're applied not re-derived. Use after a run or at session end, when the user says "remember this", "do a retrospective", "learn from…
explainer-site
Create "How X Works" deep-dive explainer websites with the curious, technical-but-accessible style of howkeyboardswork.com. Use when asked to explain how something works, create an educational single-page guide, or build an interactive explainer site on any topic.
baoyu-comic
A tool for creating educational comics from supplied text, files, links, or topics. It can cover explanations, biographies, and tutorials in different visual styles, tones, layouts, sizes, and languages.