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 agentmods add skills/mehrad-dm/mastermind/levelupnpx skills add mehrad-dm/mastermind --skill levelupgit clone --depth 1 https://github.com/mehrad-dm/mastermindWrote 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/mehrad-dm/mastermind/levelup)<a href="https://agentmods.dev/skills/mehrad-dm/mastermind/levelup"><img src="https://agentmods.dev/badge/skills/mehrad-dm/mastermind/levelup.svg" alt="Measured on agentmods" 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 | $0.00072 | $0.01100 |
| Opus 5 | $0.00036 | $0.00550 |
| Sonnet 5 | $0.00014 | $0.00220 |
| Haiku 4.5 | $0.00007 | $0.00110 |
Grade A, and why
levelup 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 4d 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 — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MasterMind: Level Up
MasterMind improves by editing its own knowledge base (its weights are fixed). This skill is the
disciplined loop that does it. Read ~/.mastermind/engineering/active-field.md first to know the active
field and its pack path (engineering/fields/<field>/).
Pick exactly one mode
Take the mode from the argument; default to capture. Do one mode per run: they touch different files under different rules, and blending them is how a lesson lands in the wrong file.
| Mode | Trigger | Read |
|---|---|---|
capture (default) |
a correction, review finding, or bug worth remembering | below. It's the whole job |
refresh |
standards may have drifted from the live ecosystem | refresh.md before writing anything: it carries the upstream-only write allowlist |
bootstrap <field> |
a new domain or stack with no pack | bootstrap.md |
Adding or rewriting a skill or agent is not a mode. It's a separate discipline: read authoring.md.
Two memory layers: the episode, then the lesson
.mastermind/journal.md, what happened (episodic). Dated one-liners appended at each verdict: the decision, the reason, the outcome. Cheap, append-only, and the project's own file.fields/<field>/lessons.md, what to do next time (semantic). Distilled from the journal.
Keeping both is what lets MasterMind say "we tried that in March and it failed because X": a lesson alone states a rule but can no longer justify it, so it gets argued with or quietly dropped. The journal is the evidence behind the rule; distil it forward and let the old entries age out.
capture (default): harvest lessons from this session/recent work
- Read the
· wrong ·lines of.mastermind/journal.mdbefore anything else (mastermind wrong-log). A miss with its catcher named is the highest-signal lesson there is, it already states the rule that was missing. Then read the rest of the journal and scan recent work for durable generalizable lessons: user corrections ("no, do X"), realcode-reviewerfindings, bugs fixed, and choices that proved right. The journal is the higher-signal source. It is what actually happened, already dated and deduplicated. Skip one-off/project-specific noise; keep only what applies to future tasks. - For each: append a one-line rule + bracketed "why" to
~/.mastermind/engineering/fields/<field>/lessons.md. Deduplicate against existing lessons. - If a lesson is a general default (not just a gotcha), promote it into
stack-defaults.mdat the right section. That's where it will actually change behavior. - Keep it tight. A lesson that isn't load-bearing is noise; keep only what earns its place.
What ships with it
4 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.
- 4d ago First seen · 73 lines · 72 tokens per session scan A eda21a5db710
levelup is a skill published in the GitHub repository mehrad-dm/mastermind (24 stars, last pushed 4d ago), licensed MIT. It adds 72 tokens to every session and 1,100 once invoked, about $0.0004 per session on Opus 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 2026-08-30.
Other skills, from other repositories
gstack-roles
Role-based AI specialist system. Defines 23 specialist roles and teaches the AI to adopt the correct role before each task phase, like a virtual engineering team.
competitive-programming-expert
Use this skill when user needs to solve competitive programming problems. Applicable to LeetCode, Codeforces, AtCoder and similar platforms. Triggers include: algorithm problem, coding challenge, time complexity analysis, data structure implementation, TLE (Time Limit Exceeded), MLE (Memory Limit Exceeded).
learning-memory
Use when capturing or restoring a learner's persistent profile to personalize teaching across sessions.
your-skill-name
State the educational problem this skill solves and the intended outcome.
challenge-generator
Use when generating personalized practice challenges calibrated to the learner's weak areas, level, and project context.
interview-mode
Use when simulating a technical interview to build confidence, expose reasoning gaps, and train structured communication under pressure.