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 skills add ZahiriNatZuke/local-delegate --skill local-delegationgit clone --depth 1 https://github.com/ZahiriNatZuke/local-delegateWrote 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/zahirinatzuke/local-delegate/local-delegation)<a href="https://agentmods.dev/skills/zahirinatzuke/local-delegate/local-delegation"><img src="https://agentmods.dev/badge/skills/zahirinatzuke/local-delegate/local-delegation/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/zahirinatzuke/local-delegate/local-delegation"><img src="https://agentmods.dev/badge/skills/zahirinatzuke/local-delegate/local-delegation.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00138 | $0.01580 |
| Opus 5.5 | $0.00055 | $0.00632 |
| Sonnet 5.5 | $0.00028 | $0.00316 |
| Haiku 4.5 | $0.00014 | $0.00158 |
Grade A, and why
local-delegation 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 today.
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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Delegación a modelos locales (MCP local-delegate)
Tienes un servidor MCP (local-delegate) que expone modelos corriendo en una GPU propia
(vía llama-swap, Ollama, LM Studio o vLLM) como herramientas texto/imagen→texto. Delegar
a ellas conserva cuota de la suscripción: la generación verbosa y los inputs grandes (o
una imagen entera) se quedan fuera de tu contexto.
Regla de oro
¿Puedo describir el paso en UNA frase con un formato de salida explícito? → sí → usa una tool
local_*. → no (necesita razonamiento, arquitectura, criterio, multi-fuente) → hazlo tú.
El mayor ahorro: path en vez de text
Para archivos/documentos grandes, pasa path (no leas el archivo tú primero). Las
tools que aceptan path leen el archivo del lado del servidor, así el contenido
completo nunca entra a tu contexto — solo recibes el resultado corto.
El coste de no delegar
Leer un archivo de 100 KB con Read cuesta ≈ 25 000 tokens de tu contexto. La misma
tarea con local_summarize(path=…) cuesta ≈ 200 tokens (solo el resumen que vuelve).
Esa diferencia es la que se pierde cada vez que se lee un archivo grande a mano en vez
de delegarlo.
Catálogo de tools
| Tool | Cuándo | Args clave | Devuelve |
|---|---|---|---|
local_summarize |
Resumir texto o archivo largo | path o text, max_words, focus (qué te interesa conservar) |
Resumen (prosa) |
local_classify |
Etiquetar en categorías fijas | text, labels[] |
Una etiqueta |
local_extract |
Sacar campos estructurados | fields[], path o text |
JSON |
local_boilerplate |
Generar código repetitivo | spec, language, target (ruta absoluta), overwrite? |
Recibo: ruta y tamaño (el código va al archivo, no a tu contexto) |
local_delegate |
Escape genérico texto→texto | task, input, output_format, model?, chunk? |
Texto |
local_lint_summary |
Resumir salida de lint/tests/CI | path o text, max_words |
Resumen agrupado por archivo |
local_commit_msg |
Mensaje de commit desde un diff | diff o path, style? |
Mensaje (revísalo siempre) |
local_translate |
Traducir texto o archivo | target_lang, text o path |
Traducción |
local_explain_code |
Explicar qué hace un código | code o path, question? |
Explicación (prosa) |
local_describe_image |
Describir una imagen o responder una pregunta sobre ella | path, question? |
Descripción (prosa) |
local_status |
Diagnóstico de solo lectura | — | Backend/catálogo/log/VRAM |
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.
- today First seen · 91 lines · 138 tokens per session scan A fc1f2a7d00f3
local-delegation is a skill published in the GitHub repository ZahiriNatZuke/local-delegate (6 stars, last pushed today), licensed MIT. It adds 138 tokens to every session and 1,580 once invoked, about $0.0006 per session on Opus 5.5. 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 2026-10-14.
Other skills, from other repositories
add-prompt
Scaffold a new MCP prompt template. Use when the user asks to add a prompt, create a reusable message template, or define a prompt for LLM interactions.
openclaw-prompt-enhancer
Remote prompt engineering MCP server with 6 tools that turn weak prompts into powerful ones. Use when: (1) 'improve this prompt' or 'make my prompt better', (2) 'score this prompt' or 'rate prompt quality', (3) 'convert prompt to Claude format' or 'translate for Gemini', (4) 'generate system prompt' or 'create…
add-prompt
Scaffold a new MCP prompt template. Use when the user asks to add a prompt, create a reusable message template, or define a prompt for LLM interactions.
add-prompt
Scaffold a new MCP prompt template. Use when the user asks to add a prompt, create a reusable message template, or define a prompt for LLM interactions.
add-prompt
Scaffold a new MCP prompt template. Use when the user asks to add a prompt, create a reusable message template, or define a prompt for LLM interactions.
opik-optimize
Improve a prompt with the Opik Agent Optimizer — resolve the prompt, a dataset, and a metric, pick the algorithm, run a bounded optimization, check the gain on held-out data, and save the winner as a new prompt version with the optimization run link. Runs via the opik-optimizer package; reads prompts and datasets via…