Kits em resumo
| kit | versão | skills | commands | agents | hooks | rules | templates | scripts |
|---|---|---|---|---|---|---|---|---|
| kit-forge | 1.4.2 | 0 | 0 | 0 | 1 | 0 | 0 | 0 |
| operator-kit | 1.6.2 | 13 | 2 | 2 | 10 | 13 | 1 | 17 |
| continuity-kit | 1.4.1 | 2 | 0 | 0 | 3 | 0 | 12 | 2 |
| lane-kit | 1.4.1 | 1 | 0 | 0 | 4 | 0 | 4 | 4 |
| health-kit | 1.3.3 | 2 | 0 | 0 | 1 | 0 | 0 | 3 |
| claude-dev-kit | 1.3.3 | 8 | 0 | 0 | 1 | 0 | 0 | 2 |
| supabase-pack | 1.1.2 | 2 | 0 | 0 | 0 | 0 | 0 | 0 |
| agent-framework-wizard | 1.2.1 | 1 | 0 | 0 | 0 | 0 | 5 | 0 |
| dev-squad-kit | 1.1.1 | 3 | 12 | 12 | 0 | 0 | 0 | 0 |
| gotcha-memory | 1.0.3 | 1 | 0 | 0 | 3 | 0 | 0 | 0 |
| total | 33 | 14 | 14 | 23 | 13 | 22 | 28 |
kit-forge 1.4.2
O gate de IP/PII + o montador de kits. ip_pii_linter + kit_assembler (com guard_origins integrado) + kit_doctor (verify/install/registry, 6 estagios: detect/prereqs/profile/configure/wire-suggest/smoke) + guard_origins + wire_settings + install_git_hook + skill_lint + tools/browse.py (menu interativo do marketplace), contrato unico de exit (0 ok/no-op, 1 warn, 2 block, 3 erro). Sem --skip-lint.
Hooks
| evento | script |
|---|---|
| (wiring pelo instalador) | guard_origins.py |
operator-kit 1.6.2
Módulo operacional do harness: gates executáveis, loops governados, pass@k/pass^k standalone, política audit/enforce, preflight, planejamento spec-driven e dois checkers read-only.
Skills
| skill | o que faz |
|---|---|
adversarial-refuter | Before accepting "done/ready", dispatches read-only refuters that try to KNOCK DOWN the claim against the disk/live source |
claude-md-from-profile | Generates the project's CLAUDE.md block FROM operator-profile.yaml - floor first (what the AI does not decide), then where things live, the rules of done and the flow. Idempotent, signed; refuses to overwrite a hand-edited block |
delegate-with-handback | Delegates a long/independent task to a 2nd agent with explicit context and a verification gate on the return |
doc-consolidator-dedup | Merges N overlapping docs/plans into a single deduplicated work-list, archives the superseded ones with a stub-redirect — grep/ls before creating/classifying (LC-3) |
dual-report-builder | Generates TWO versions of the same analysis/report — INTERNAL (raw, failures exposed, dark) and EXTERNAL (premium, positive, no failures exposed, light) — as self-contained HTML with pure-CSS charts, print-friendly. The external version goes through a sober-register gate (banned phrases from the profile). Use when producing a report/dashboard with a dual audience (internal team + client/stakeholder). |
gate-sheet-collector | Drains everything that depends on the human into ONE form (exact command + what-it-unblocks), without ever blocking the loop |
gated-improvement-proposal | Every self-edit of the harness (rule/CLAUDE.md/prompt/hook) becomes a PROPOSAL that passes a gate before being applied — never a sensitive auto-merge |
live-source-prover | Before citing ANY number/status/metric, re-derives it at the live source and labels it "live @ HH:MM + source" — never repeats stale data |
parallel-dispatch | Fires independent tasks in waves with a concurrency ceiling + sequential fallback on rate limit |
pre-clear-boot-block | Before a /clear, emits the DONE / MISSING / read-in-this-order block so the fresh session resumes with zero loss |
ralph-loop-driver | Turns the agent into an autonomous lead engineer — reads charter+work-list, executes until exhausted, self-prompts, stops at the stop-conditions |
rls-audit | Really audits the RLS of a Supabase project — pg_policies per permissive anon policy + get_advisors, not just the relrowsecurity flag |
supabase-edge-scaffold | Scaffolds a Supabase Edge Function with correct CORS + service-role + error handling, instead of copying boilerplate by hand |
Commands
/cancel-ralph-gate /ralph-gate
Agents
refutador silent-failure-hunter
Hooks
| evento | script |
|---|---|
PreToolUse · Bash | operation_guard_portable.py |
PreToolUse · Bash | snapshot_rollback_gate.py |
PreToolUse · Bash | external_send_draft_gate.py |
PreToolUse · Bash | fact_force_gate.py |
PreToolUse · Edit|Write|MultiEdit|NotebookEdit | secret_scan_on_write.py |
PreToolUse · Edit|Write|MultiEdit|NotebookEdit | project_root_confirm.py |
PreToolUse · Edit|Write|MultiEdit|NotebookEdit | fact_force_gate.py |
UserPromptSubmit · * | rule_capture.py |
Stop · * | ralph_gate.py |
Stop · * | autoprompt_resume.py |
Rules
agent-cognition agent-integrity epistemic-standards gateguard learned-corrections loop-cost-budget loop-maker-checker loop-operator loop-passk loop-patterns-catalog no-secrets-in-memory partial-autonomy-slider stale-replay-guard
Templates
loop-charter-template.md
Scripts
audit_plan.py claude_md_from_profile.py debt_ledger.py delta_inventory.py determinism_harness.py distill_corrections.py done_gate.py drift_check.py gate_sheet_panel.py goal_ledger.py goal_review.py health_probe.py live_count.py passk_eval.py preflight.py status_now.py verify_ladder.py
Documentos e registros
docs/HOOK-SKILL-STANDARDS.md docs/HOOK-SKILL-STANDARDS.pt-BR.md docs/MCP-RUNBOOK.md docs/MCP-RUNBOOK.pt-BR.md docs/RULES-EAGER-BUDGET.md docs/RULES-EAGER-BUDGET.pt-BR.md
continuity-kit 1.4.1
Handoff-v1.1: uma sessao sobrevive a parada/clear/crash sem perder o proximo passo. Schema com git-block + re_derive_cmd (todo numero carrega o comando que o re-deriva ao vivo) + verify_first_cmd (conferir ao vivo antes de repetir qualquer acao que o handoff descreve), hooks Stop/PreCompact/SessionStart. Inclui doc-rollup (historico/evolucao com degradacao embutida) + pre-clear (longo-prazo + curto-prazo).
Skills
| skill | o que faz |
|---|---|
doc-rollup | Keeps the project's history/evolution docs (changelog, narrative timeline, state snapshot, lessons, session wrapup) up to date after a significant session — with built-in degradation (a stamp instead of endless narrative) from day 1. |
pre-clear | Before a /clear, consolidates the LONG TERM (conditional doc-rollup — how we got here) and the SHORT TERM (handoff — what comes next), then renders the BOOT BUNDLE. |
Hooks
| evento | script |
|---|---|
SessionStart · * | handoff_inject.py |
Stop · * | handoff_guard.py |
PreCompact · * | handoff_guard.py |
Templates
00-DEPLOY.template.md 00-ISOLATION-AND-RECOVERY.template.md 00-PROCESSES.template.md 00-READ-FIRST.template.md 00-ROLLBACK.template.md 00-STATE.template.md 00-VISION.template.md LEARNINGS.template.md loop-charter.template.md settings-continuidade.template.json wave-prd.template.md wave-review.template.md
Scripts
doc_rollup.py state_mirror.py
lane-kit 1.4.1
N sessoes sem colisao. Lane board, maker!=checker cross-model, lock por diretorio, git-guard e territory-guard. O checker_router detecta Codex, Cursor e Gemini e escolhe um provider diferente do maker. Best-of-N: compete/select registram qual de N tentativas concorrentes venceu.
Skills
| skill | o que faz |
|---|---|
lane-coordinator | Coordinates N concurrent sessions (lanes) over the same repo through a whiteboard with a state machine (lane_board.py) — from CLAIMED to MERGED, with cross-model maker≠checker enforced in code, not by textual discipline. |
Hooks
| evento | script |
|---|---|
SessionStart · * | lane_register.py |
PreToolUse · Bash | lane_git_guard.py |
PreToolUse · Edit|Write | lane_territory_guard.py |
PostToolUse · * | lane_register.py --heartbeat |
Templates
lane-registry.example.json lanes.example.yaml REORIENT-MAILBOX.template.md status-stakeholder.template.html
Scripts
checker_router.py lane_board.py lane_effects.py lane_rescue.py
health-kit 1.3.3
Sonda de saude de servicos (http/cmd) config-driven por profile.yaml + segmento de statusline com detalhe por-servico (api:OK db:DOWN), cache-first (statusline nunca toca rede). Doutrina embarcada: health de SERVICO != health de DADO. +dashboard-builder (Grafana/SigNoz).
Skills
| skill | o que faz |
|---|---|
dashboard-builder | Builds monitoring dashboards (Grafana, SigNoz and similar) that answer real operator questions, not "show every metric that exists". Use when turning a list of metrics into a genuinely operable dashboard. |
health-check | Probes a config-driven list of services (HTTP or local command) and writes a JSON cache that another tool (e.g. the statusline) can read without touching the network — never on the hot path itself, it only generates the cache. |
Hooks
| evento | script |
|---|---|
SessionStart · * | pyrun.sh "${CLAUDE_PLUGIN_ROOT}/scripts/health_probe.py" --quiet |
Scripts
gate_sheet_panel.py health_probe.py wire_statusline.py
claude-dev-kit 1.3.3
Ferramentas de construir ferramentas: skill-writer, hookify, plugin-dev, teaching, wiring reversivel, secret scan, architecture-decision-records, search-first e skill-scout.
Skills
| skill | o que faz |
|---|---|
architecture-decision-records | Captures architectural decisions made during the session as structured ADRs (context, alternatives considered, consequences) in docs/adr/. Use when the user decides between significant alternatives (framework, database, pattern) or asks "why did we choose X?". |
claude-dev-setup | Installs the base Claude Code hooks in a new project - idempotent and reversible settings wiring (never overwrites someone else's config without --force) + chain-preserving git hook (never replaces an existing pre-commit hook). |
hookify | Creates REAL hooks for Claude Code — executable scripts (stdin JSON, exit 0/1/2), registered via a plugin's hooks.json or pasted into settings.json. Use when the user wants to create a hook, a safety rule, a custom validation, a lifecycle hook. |
plugin-dev | Packages skills/hooks/commands into an installable Claude Code plugin — the real anatomy (.claude-plugin/plugin.json + hooks/hooks.json + ${CLAUDE_PLUGIN_ROOT}), no framework/build step. Use when the user wants to create a plugin, package an extension, or distribute a set of skills/hooks. |
search-first | Searches for an existing library/tool/pattern BEFORE writing new code - covers package registries (npm/PyPI), MCP and GitHub, in addition to the local grep. Use before creating a utility, helper, or new integration. |
skill-scout | Searches for skills locally, in the marketplace, on GitHub and on the web BEFORE creating a new skill — avoids duplicating work that already exists. Use when the user says "create a skill", "is there a skill for X?", or you are about to suggest creating a new skill. |
skill-writer | Guides the creation of Agent Skills for Claude Code — structure, frontmatter, effective descriptions, and validation against the SKILL-CONTRACT. Use when the user wants to create, write or structure a new skill. |
teaching | Turns any technical output (creation, structure, architectural decision) into a learning opportunity - a tree of where the element lives, an x-ray of what-it-is/where-it-sits/what-it-is-for, a connection map, a business analogy, and explained decisions. Always use in technical output for a non-programmer reader. |
Hooks
| evento | script |
|---|---|
PreToolUse · Edit|Write|MultiEdit | secret_scan_on_write.py |
Scripts
install_git_hook.py wire_settings.py
Documentos e registros
docs/hook-template.py docs/SKILL-CONTRACT.md docs/SKILL-CONTRACT.pt-BR.md docs/skill-template.md docs/skill-template.pt-BR.md
supabase-pack 1.1.2
rls-audit (RLS de verdade via pg_policies + get_advisors) + supabase-edge-scaffold (Edge Function TS com deno check).
Skills
| skill | o que faz |
|---|---|
rls-audit | Audits a Supabase project's RLS for real — pg_policies per permissive anon policy + get_advisors, not only the relrowsecurity flag |
supabase-edge-scaffold | Scaffold of a Supabase Edge Function with correct CORS + service-role + error handling, instead of copying boilerplate by hand |
agent-framework-wizard 1.2.1
Wizard de 6 passos (check_python->check_git->check_deps->configure->validate->generate_and_summary) para gerar o esqueleto de um agente/skill novo. --interview/--answers nao-interativos + --demo; skip-exists com --force.
Skills
| skill | o que faz |
|---|---|
agent-framework-scaffold | Runs the 6-step wizard that generates the skeleton of a new project (operator-profile.yaml + the chosen templates) — check environment → configure → validate → generate. |
Templates
00-PROCESSES.template.md 00-READ-FIRST.template.md 00-STATE.template.md 00-VISION.template.md agents
dev-squad-kit 1.1.1
Squad com 12 papéis disponíveis como slash commands e subagents reais, tools explícitos e QA read-only, mais 3 skills de leitura/consolidação paralela token-safe. Não traz árvore de tasks/templates: os comandos de papel que a esperam (*create, *task, *workflow) precisam da sua.
Skills
| skill | o que faz |
|---|---|
pp-consolidate | Parallel Process Consolidate - consolidates the outputs of parallel sessions/agents into a single, deduplicated, verifiable verdict |
pp-discovery | Parallel Process Discovery - token-safe inventory of repositories, folders and large artifacts before deep analysis |
pp-xray | Parallel Process X-Ray - line-by-line reading of an external module/repo with evidence, risks and calls for the next investigation |
Commands
/analyst /architect /data-engineer /dev /devops /master /pm /po /qa /sm /squad-creator /ux-design-expert
Agents
analyst architect data-engineer dev devops master pm po qa sm squad-creator ux-design-expert
gotcha-memory 1.0.3
Loop de aprendizado operacional standalone: a falha vira conhecimento. Depois de cada comando Bash que falha, o postflight registra o evento classificado por familia de erro; quando o mesmo tipo recorre N vezes numa janela, vira um GOTCHA -- uma licao acionavel que o preflight injeta ANTES da proxima execucao da mesma tarefa. Gotchas curated (suas regras) sao always-on. Deteccao conservadora: so sinal claro de erro conta, ambiguo nao e falha. Dois hooks WARN-only (exit 0 sempre) -- o loop de aprendizado jamais bloqueia o fluxo. stdlib only.
Skills
| skill | o que faz |
|---|---|
gotcha-memory | Operational learning loop — records command failures, detects recurrence and injects the lesson as a preamble before the next execution of the same task. Use to query/seed/debug the project's gotcha memory. |
Hooks
| evento | script |
|---|---|
PreToolUse · Bash | gotcha_preflight.py |
PostToolUse · Bash | gotcha_postflight.py |
PostToolUseFailure · Bash | gotcha_postflight.py |