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.
git clone --depth 1 https://github.com/pedrol-cmd/brain-drinWrote 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/commands/pedrol-cmd/brain-drin/self-improve)<a href="https://agentmods.dev/commands/pedrol-cmd/brain-drin/self-improve"><img src="https://agentmods.dev/badge/commands/pedrol-cmd/brain-drin/self-improve.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 5.1 | $0.00023 | $0.00644 |
| Opus 5 | $0.00012 | $0.00322 |
| Sonnet 5 | $0.00005 | $0.00129 |
| Haiku 4.5 | $0.00002 | $0.00064 |
Grade A, and why
self-improve 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 — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/self-improve
Princípio
Sistema que aprende tudo fica gordo. A regra é: só entra o que prova valor. Tudo que não prova, sai.
Resultado esperado: CLAUDE.md e agentes menores e mais precisos a cada ciclo, não maiores.
Coleta de evidências
Ler .drin/memory-compiler/daily/ (últimas 2 semanas) e .drin/memory-compiler/knowledge/.
Separar em dois grupos:
SINAL (vale aprender):
- Padrão apareceu 3+ vezes nas sessões
- Pedro disse "exatamente isso", "perfeito", "assim sempre" — validação explícita
- Abordagem resolveu problema que antes travava
- Pedro repetiu o mesmo pedido de formas diferentes — indica lacuna no sistema
RUÍDO (ignorar):
- Aconteceu 1 vez, contexto específico
- Pedro corrigiu depois — aprender o erro seria pior
- Pedro pediu para cortar, simplificar, parar de fazer algo
- Específico do projeto atual, não universal
Análise do sistema atual
Para cada item no CLAUDE.md, agentes e comandos, verificar:
| Pergunta | Se não → |
|---|---|
| Isso foi usado nas últimas 2 semanas? | Candidato a remoção |
| Isso está gerando o comportamento certo? | Candidato a reformulação |
| Isso está duplicado em outro lugar? | Consolidar ou remover |
| Isso está fazendo o sistema maior sem torná-lo melhor? | Remover |
Propostas — formato obrigatório
Cada proposta deve ter:
TIPO: ADICIONAR / REMOVER / REFORMULAR
ONDE: [arquivo exato, linha se aplicável]
O QUE: [mudança específica]
EVIDÊNCIA: [observado X vezes em [datas] / validado explicitamente em [data]]
IMPACTO ESPERADO: [por que isso torna o sistema melhor]
Nenhuma proposta sem evidência. "Acho que seria melhor" não é evidência.
Regras de qualidade
- Redução > Adição. Se o ciclo não removeu nada, o sistema provavelmente só inchou.
- CLAUDE.md máximo 80 linhas. Se ultrapassar, algo menos valioso precisa sair.
- Agente novo só entra se substituir comportamento manual recorrente — não por "pode ser útil".
- Comando novo só entra se Pedro usaria pelo menos 1x/semana.
- Toda adição tem data de revisão — se em 30 dias não foi usado, remove automaticamente.
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 · 78 lines · 23 tokens per session scan A a4226fbb4fa4
self-improve is a command published in the GitHub repository pedrol-cmd/brain-drin (11 stars, last pushed 5mo ago), licensed MIT. It adds 23 tokens to every session and 644 once invoked, about $0.0001 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 2026-09-03.
Other commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.