zedarvates

monte_cristo

Independent strategic outsider above the blue and red teams. Use for architecture resets, high-cost research decisions, inherited assumptions, and stalled adversarial reviews; do not use for routine review or autonomous mutation.

cluster

Treat the homelab/micro-cluster as one schedulable resource — discover every reachable machine and spread cheap work across them (least-recently-used first) so idle boxes get the next task, recovering wasted local capacity instead of paying the cloud. Use to see cluster status, route a task to the idlest machine, or hand a task to a trusted agent on another machine. Also use when the user mentions homelab, cluster, idle machines, load balancing, or distributing work across machines.

conductor

Route a high-level goal to an ordered, local-first plan of capabilities — the generalised router — and optionally EXECUTE the plan's read-only steps. Reads the capability registry/curator, composes steps ordered by the system's layers (SENSE→DECIDE→ACT→REMEMBER→GOVERN→DEPLOY), annotates each local vs cloud with a concrete command, and estimates the goal's effort. The executor runs the safe analysis steps unattended and confirm-gates anything that mutates state or escalates to the cloud. Use when the user states a goal and you need to decide which botte capabilities to use in what order (or run them), or asks "how should I approach X with this toolkit".

demo

Live ANSI dashboard of the belt's decisions — routing, token savings, micro-NN outputs, escalations, cache hits — either a built-in scripted scenario (no LLM, no network, works on a bare machine) or tailing a real project's event log. Use when the user wants to see/demo what the routing belt is doing, record a README GIF, or watch live decisions while an agent works.

fast_context

FastContext Agent — exploration repo déterministe. Parse une requête d'exploration → READ/GLOB/GREP ciblés → rapport compact (fichier:ligne:score). Remplace 56% des appels LLM de type "read/search" par des opérations stdio à ~5ms, 0 token. Use when the agent needs to understand a codebase, find patterns, locate imports, or gather context without an LLM call.

meta_harness

Meta-Harness orchestre les skills Botte comme des agents interchangeables dans un pipeline gouverné. Planifie → exécute en sandbox → review croisé → applique avec garde-fous. Inspiré d'Omnigent mais 100% Botte-native.

events

Append-only JSONL decision log (.botte/events.jsonl) that every filter in the belt writes to — routing, cache hits, escalations, micro-NN outputs. The single source of truth demo mode, the live dashboard, and session replay all read from. Use when you want to see or emit a live feed of routing/cache/escalation decisions, or when building a tool that needs to watch the belt work in real time.

auto-memory

Memory as a learnable skill — store, recall, compress, and consolidate agent memories. Inspired by Stanford AutoMem.

bench

Reproducible token/cost benchmark — runs a fixed task corpus through the real auto_router decision logic and compares it against a "no routing, everything to cloud STANDARD" baseline. Turns the README's savings claim into a checkable number instead of an assertion. Use when the user wants proof of token savings, a benchmark for a PR/README, or numbers to back an integration pitch (e.g. Hermes).

capabilities

The system's self-model — a capability registry (scans every SKILL.md into a layered tree SENSE→DECIDE→ACT→REMEMBER→GOVERN→DEPLOY) plus a "curator" that picks the right capabilities for a goal locally. Use to see the whole toolkit as a system/arborescence, to let an agent discover what botte-secrète can do, or as the data the Conductor reads to compose a plan.

control_loop

Close the system into a self-improving loop — measure routing outcomes (local %, token savings, escalation/success rates) and adapt the effort→tier thresholds the auto-router reads, so it gets better at keeping work local without failing. Use to see routing stats, tune the router from real outcomes, or reset thresholds. Also use when the user wants the system to adapt/learn its routing automatically.

cost_estimator

Estimate what a task or a fix will cost — tokens, model/tier, money ($), and wall-time — using the tiered cost model. Use whenever the user wants to know the cost of a correction, audit, or task before running it, or to compare local vs cloud cost.

decision-ladder

Ponytail-inspired YAGNI enforcement — climb the decision ladder before writing any code. stdlib → regex → existing module → new code.

fix

List a project's correctable issues — confirmed dead code, duplication, stale directive references — each with a tokens·model·money·time cost estimate and a total. Plan-only by design (never edits code automatically). Use when the user asks what's worth fixing and what each fix costs.

mcp_gateway

MCP Gateway — expose toutes les skills Botte comme outils MCP. Découverte automatique, schémas d'entrée, transport stdio. Compatible Claude Code, Codex, Cursor, et tout client MCP.

app_test

Local-first GUI/app testing by image matching (SikuliX) — turn a small JSON spec referencing your button images into a runnable SikuliX script and run it locally, using a vision NPU (Hailo-8/10) or local vision model instead of cloud vision. Use when the user wants to test a desktop/game/web app "for real" by clicking buttons, has button images already, or mentions SikuliX, image-matching tests, or local UI tests.

checkup

Run the canonical, already-optimal project checkup in one command — policy presence, directives health, per-component metrics, infra tips, duplication, and drift detection — so you never have to hand-write a good checkup prompt. Use when the user says "do a complete checkup", "botte doctor", after a component update, when onboarding to a project, or when multiple agents/devs may have caused drift.

directives_audit

Audit a project's AI-agent guidance files — CLAUDE.md, AGENTS.md, .cursorrules, copilot-instructions.md, GEMINI.md, intent docs and specs, in markdown, text or HTML. Use when checking whether a repo has agent instructions, whether they are healthy/consistent, or before an agent starts work on an unfamiliar codebase. Also use when the user mentions CLAUDE.md, AGENTS.md, agent rules, "does this project have instructions", or stale/oversized directive files.

docs_steward

Scoped documentation map for multi-component projects (server + client + tools + …). Detects components, classifies every doc as global vs component-scoped, and produces a per-component index (DOCS.md) listing local docs + links to the relevant global docs — so an LLM coder bounded to one component loads only its scope, not every other component's documentation. Frames token cost (full project docs vs scoped load) and treats .md as LLM-facing, .html as human reference. Use when a project has several components and you want to cut the docs an agent must read, or asks how to organise docs for a monorepo.

llm_backends

Discover, audit and use local LLM servers (LM Studio, Ollama, LocalAI, vLLM, llama.cpp) on this machine or the network to offload work from the cloud and save tokens. Use when the user mentions local models, LM Studio, Ollama, "run it locally", token savings via local hardware, or wants to know what models their machine can run.

dashboard

Generate one self-contained, timestamped HTML dashboard of the system's cost picture — routing savings (control loop), metric trends, current metrics, and the cost of outstanding fixes. Also renders as a live ANSI terminal view (--tui, --watch) and serves a live HTTP API (api.py). Use when the user wants a single visual view of cost/savings/health over time, or a live terminal view they don't have to open a browser for.

hermes-bridge

Expose routing, local chat, fusion, skill search, infrastructure advice, and privacy-safe QA run manifests to Hermes-Agent or another OpenAI-function-calling framework, plus a one-call MCP config generator. Use when connecting botte-secrète's routing and evidence contracts to another agent framework.

auto_router

Auto-decide whether a task runs on a LOCAL model or a CLOUD model (DeepSeek, GLM, Nemotron, Grok, Gemma, …) from an automatic effort estimate, and run multi-model fusion (cascade, draft→refine, vote). Use when the user wants automatic local-vs-cloud routing, to add cloud LLM providers, to make local and cloud models collaborate, or mentions effort-based routing, model fusion/ensemble, OpenRouter, DeepSeek, GLM, Nemotron, or Grok.

llm_mcp

MCP server that lets Claude Code (or any MCP client) discover and call local LLM servers (LM Studio, Ollama, …) as tools, to offload cheap tasks off the cloud. Use when the user wants to wire local models into their agent, register an MCP server, or have the agent automatically route simple tasks to local hardware.

agent-cache

Cache les réponses des agents pour skipper l'exécution quand l'output est prédictible. Use when you want to cut 10-15% by avoiding redundant agent runs.

bootstrap

Deploy Botte Secrète's token-saving stack into a target project — wire the botte-llm MCP server into .mcp.json, audit the project's agent directives, and write a .botte config + setup report. Use when the user wants to "install botte", "set up token savings on a project", reduce an existing project's token/cost usage, or onboard a repo to local-first routing. This is the capstone that makes the toolkit actually save money on real projects.

botte-proxy

Transparent LLM compression proxy for botte-secrete — sit between any AI agent and its LLM API to compress requests by 40-95%. Use when you want token savings without changing agent code.

context_budget

Pick the optimal set of skills/docs to load for a task under a token budget — an exact 0/1 knapsack (maximize relevance while summed token cost stays under budget), not an LLM "decide what's relevant" call. Deterministic, stdlib, 0 cloud tokens. Cuts the always-on context cost (loading the whole skill catalog every turn). Use when you want to load only the most relevant context within a budget, or to decide which skills/docs an agent should read for a task.

context_profiler

Measure a project's always-on prefix (agent directives + core rules + MCP tool schemas + skill catalogue) in tokens and as a % of small local-model windows (64k/128k/256k), with a concrete reduction plan (lazy tool loading, on-demand skill search). Use to see how much of a modest machine's usable context is spent before any real work, and how to shrink it so weaker machines can run local LLMs usably.

cwe_kb

Local CWE knowledge base (RAG) to enrich and de-noise security findings — match a finding (or any text) to the relevant weakness by exact id or local-embedding similarity, and attach the weakness name, description, and concrete mitigation. Deterministic, offline, 0 cloud tokens. Use to explain a CWE, find the likely weakness for a code snippet, or enrich taint/security findings with "why + how to fix".

docgen

Generate documentation with a local model drafting and the cloud only refining (0 cloud tokens for the draft), plus a local session review that summarises what a work session did. Use to write README/module docs/CHANGELOG/guides/ADRs cheaply, or to produce end-of-session notes/learnings from a transcript.

infra_advisor

Audit the local cluster's hardware/software/MCP setup and recommend changes that cut token cost — GPU upgrades, Hailo NPU for vision, moving the inference node to Linux, running Qdrant locally, wiring MCP — with an ASCII cluster diagram. Also provides an auto one-pass audit (directives + infra + code duplication + skills) on the project. Use when the user asks how to reduce token/usage cost via hardware/infra, wants cluster setup tips, an ASCII diagram of their setup, or a quick all-in-one audit.

ingest

Local-first web scraping and source ingestion — fetch a URL, extract clean text locally (0 cloud tokens), optionally structure it with a local model, and store it in a Qdrant collection (the second-brain "foundation") for later recall. Use for /web-scraping, /ingest-source, building a knowledge foundation, or recalling past ingested sources. Web fetch + Qdrant over stdlib HTTP.

metrics

Cost-focused project metrics, broken down per component — LOC by language and component, duplicate-function groups, directive health, always-on context cost (CLAUDE.md tokens × turns), local-routing posture, skill-search tokens avoided, and the audit's own (near-zero) cost. Use when the user wants to quantify a project's token/cost profile, see LOC/health per component, or measure what the toolkit saves.

nlp_deterministic

Classify and extract from text WITHOUT an LLM — intent classification (lexical overlap + local embedding), entity extraction (regex/gazetteers for urls/emails/ips/paths/env vars/flags/numbers), and stopword-filtered keyword frequency. Deterministic, instant, 0 cloud tokens. Use instead of asking a model to classify text or extract entities, and as the routing/intent layer that keeps cheap language decisions off the LLM.

nn_audit

Audit the micro-NNs — is each model grounded in REAL data, or a synthetic copy of a hand-coded rule? Scans skills/botte_nn and reports, per model, the training data source (real/synthetic/unknown), whether the model file records provenance (trained_on/eval_accuracy), whether a test guards a real-world output, and a grounded/synthetic verdict. Deterministic, 0 cloud tokens. Use to tell which learned components are real vs placeholder, and which should be grounded or replaced by the rule they imitate.