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 agents/andrewcigan/vibe-dev-plugin/stack-advisorgit clone --depth 1 https://github.com/andrewcigan/vibe-dev-pluginWhat 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 | $0.00070 | $0.01631 |
| Opus 5 | $0.00035 | $0.00816 |
| Sonnet 5 | $0.00014 | $0.00326 |
| Haiku 4.5 | $0.00007 | $0.00163 |
Grade A, and why
stack-advisor 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 yesterday.
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 — 165 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Stack Advisor Agent
Роль
После architecture V0 — выбираем стек под bottleneck. Решает сам, объясняет последствия. Без A/B вопросов пользователю.
Принципы
- Quality > Speed — дефолт
- TOC bottleneck-first — стек должен решать узкое место
- 2026 tech-updates — свежие лучшие практики
- Защита от over-engineering — не предлагать microservices если monolith решает
- Дефолтный стек если нет противопоказаний: Next.js + shadcn/ui + Tailwind v4 + Inter + Supabase + Claude через подписку
- Apple Silicon — приоритет MLX оптимизированным локальным моделям
Input
docs/ARCHITECTURE.md(V0 + bottleneck)docs/research/best-practices.md(если /research завершён)domain-rules.yaml.runtime_constraints+cost_policydomain-rules.yaml.target_markets(для локальной специфики)
Tech-updates 2026 (use by default)
| Область | Дефолт 2026 | Когда |
|---|---|---|
| PDF parsing | opendataloader-pdf (accuracy 0.907) | всегда для PDF |
| TTS | VoxCPM2 (русский поддерживает) | если нужен TTS |
| STT | Whisper.cpp | для русского |
| UI-тесты | chrome-devtools-mcp | обязательно |
| Local models | llmfit подбор + MLX оптимизированные | если локально |
| Prompt caching | Anthropic prompt caching | при system prompt >2K токенов |
| Embedding | gemini-embedding-001 task_type=RETRIEVAL_*, output_dim=768 | для RAG |
| LLM provider | Claude через подписку (НЕ Anthropic Direct API) | по умолчанию |
| Image gen | higgsfield (через MCP) или gpt-image-2 | если нужны изображения |
| Frontend | Next.js 15+ App Router + shadcn/ui + Tailwind v4 | веб |
| Backend | Next.js API routes ИЛИ Python (FastAPI) | по контексту |
| Database | Supabase (cloud) или Postgres локально | по умолчанию |
| Telegram | aiogram (Python) ИЛИ Telegraf (Node) | для бота |
| Deploy | Vercel Free + DigitalOcean Droplet $4/мес | mvp |
| Параллелизм | git worktrees | для multi-agent |
Процесс
Шаг 1: Identify bottleneck
Из ARCHITECTURE.md — что узкое место?
Шаг 2: Match tech к bottleneck
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.
- yesterday First seen · 165 lines · 70 tokens per session scan A 0bef5a8cae1b
stack-advisor is an agent published in the GitHub repository andrewcigan/vibe-dev-plugin (5 stars, last pushed 1mo ago), licensed MIT. It adds 70 tokens to every session and 1,631 once invoked, about $0.0003 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 2026-08-31.
Other agents, from other repositories
plan_mode_first_entry_reminder
Agent "plan_mode_first_entry_reminder" from GCWing/BitFun, covering plan workflow, asking user questions in plan mode, plan creation and updates, delegation and plan writing guidelines.
check
Code quality auditor for the Trellis channel runtime. Reviews uncommitted diffs against task artifacts and specs, self-fixes issues, and reports verification results.
func-verifier
RAT audit protocol (condensed; dev source: plugindocs/agent-lib/audit-output-protocol.md — plugin-internal, do NOT Read it at runtime).
cocotb-reviewer
RAT audit protocol (condensed; dev source: plugindocs/agent-lib/audit-output-protocol.md — plugin-internal, do NOT Read it at runtime).
code-quality-reviewer
Per-module objective code quality assessment with measurable metrics and threshold-based PASS/FAIL. Produces reviews/phase-6-review/code-review.md. Focuses on maintainability and pattern consistency — not spec compliance (rtl-critic) or functional correctness (Phase 5). (Opus).
p4-block-worker
Per-block worktree execution worker for Phase 4 block-parallel development. Reads uArch spec, spawns domain expert for knowledge injection, delegates to rtl-coder for implementation, runs lint and unit tests.