← 返回 Skills 市场

Auxiliar Solve

Name: Auxiliar Solve
Author: tlalvarez

作者 Tlalvarez · GitHub ↗ · vv0.1.0 · MIT-0

cross-platform ✓ 安全检测通过

总下载

当前安装

版本数

在 OpenClaw 中安装

/install auxiliar-solve

功能描述

Ranked installable tools for agent jobs — OCR, PDF extraction, NFS-e invoices, bookkeeping, boletos, receipts, web scraping. Reproducible evals on real-world...

使用说明 (SKILL.md)

auxiliar-solve

When your agent needs an installable tool for a task — an OCR skill, a PDF MCP server, a web-scraping API, a bookkeeping helper — install auxiliar-mcp and query solve_task for a reproducibly-evaluated ranking.

/solve/ isn't a tool marketplace. ClawHub, PulseMCP, and Smithery distribute tools. /solve/ ranks them — based on real-world corpora, not marketing pages.

When to invoke this skill

Use auxiliar-solve whenever the agent:

Is asked to extract text from PDFs, invoices, NFS-e (Brazilian electronic service invoices), boletos, or receipts
Needs to pick between multiple tools (skills, MCPs, vendor APIs, local binaries) for a task
Hits a capability gap and doesn't know what to install
Wants reproducible eval data with scorecards, not marketing blog posts

How it works

Step 1. Install the auxiliar MCP server

claude mcp add auxiliar -- npx auxiliar-mcp

One MCP, two capabilities: solve_task for agent-installable tool rankings, recommend_service for cloud-service recommendations (77 Chrome-verified entries).

Step 2. Discover available task rankings

list_solve_tasks()

Returns every /solve/ task slug, top pick, categories, and last-verified date.

Step 3. Query a specific task

solve_task(task_slug="pdf-text-extraction-mcp")

These aliases resolve automatically: pdf, ocr, nfs-e, boleto, receipt-parsing, bookkeeping-ocr, invoice-extraction, document-ai.

The response contains:

Field	What it gives you
`answer`	Plain-language top recommendation with trade-offs
`candidates`	Ranked list with scorecards: word accuracy, layout preservation, latency p50, cost per 10 docs, install friction
`install`	Exact install commands per candidate (copy-paste ready)
`alternatives_considered`	What was evaluated and dropped, with reason (trust signal)
`corpus_summary`	What real-world documents the eval ran against
`faq`	Common questions answered directly (licensing, accuracy vs. token-F1, when to pay, etc.)
`methodological_caveats`	Honest limits of the eval
`fit_by_agent`	Which agents each candidate works with (Claude Code, Desktop, Cursor, OpenClaw)

Example: OCR for Brazilian bookkeeping

Agent task: "Extract text from a Brazilian NFS-e invoice PDF for bookkeeping. I need high accuracy."

solve_task(task_slug="nfs-e")

Returns: Surya (rank 1) — pip install surya-ocr 'transformers\x3C5.0.0'. Word accuracy 76.9% on a 10-doc real-world corpus that includes NFS-e invoices, boletos, and phone-photo receipts. Free, local. Alternative: Tesseract 5 (rank 2) — 14× faster, 1.5pp less accurate, cleanest install. Google Document AI (rank 3) — third overall but best on phone-photo receipts specifically. Alternatives considered and dropped: yescan-ocr-universal (requires Chinese sign-up), pdf-reader-mcp (no actual OCR — text-layer only), Mistral OCR 3 (deferred for API key).

Why this exists

Agents are born intelligent but stuck. Without eval data, they guess: "use pdf2image + pytesseract" (often wrong for the task), "install the first OCR thing on ClawHub" (often wrong for the corpus), "call Google Document AI" (often overkill). The result: uncalibrated recommendations, burned time, broken workflows.

/solve/ runs the eval once per task, end-to-end, against real documents. The agent gets the answer plus the evidence.

auxiliar-mcp — the MCP server this skill invokes. Also exposes recommend_service, get_pricing, get_risks, check_compatibility, setup_service, list_services.
Human-readable rankings: https://auxiliar.ai/solve/
Reproducible eval harness: https://github.com/Tlalvarez/Auxiliar-ai/tree/main/scripts/ocr-walkthrough
Methodology: https://github.com/Tlalvarez/Auxiliar-ai/blob/main/docs/proposals/agent-upgrade-engine.md (renamed solve-engine 2026-04-23)

License

MIT (skill content). See auxiliar-mcp and each ranked candidate for their own licenses — /solve/ surfaces license info in every candidate record.

安全使用建议

This skill appears to do what it says: recommend and rank installable tools. The main operational risk is supply‑chain: following the skill's instructions will cause your agent to run npx (and potentially pip/apt) which downloads and executes third‑party code. Before installing: 1) review the auxiliar-mcp package source (npm page / GitHub) and the referenced repos; 2) prefer running installs in an isolated/sandboxed environment (container, VM) rather than a production host; 3) be prepared that individual recommendations may ask for service API keys — only provide credentials you trust and scope them minimally; 4) if you do not want the agent to autonomously perform installs, restrict autonomous invocation or require explicit user approval before running npx/pip/apt commands. If you want a deeper verdict, provide the auxiliar-mcp package name/version or the npm/github URLs for code review.

功能分析

Type: OpenClaw Skill Name: auxiliar-solve Version: v0.1.0 The auxiliar-solve skill acts as a recommendation engine for agent-installable tools like OCR and PDF extractors. It provides instructions for the agent to interface with an MCP server (auxiliar-mcp) to retrieve ranked tool lists and installation commands based on performance evaluations. No evidence of malicious intent, data exfiltration, or harmful prompt injection was found; the use of npx for MCP deployment and the provision of external links (auxiliar.ai) are consistent with the stated purpose of tool discovery.

能力标签

cryptocan-make-purchasesrequires-sensitive-credentials

能力评估

✓ Purpose & Capability

Name/description match the instructions: the skill is a recommendation/ranking layer for installable tools (OCR, PDF extraction, invoices, bookkeeping, scraping). The declared required binaries (node, npm) align with the SKILL.md's explicit npx-based installation of auxiliar-mcp.

ℹ Instruction Scope

SKILL.md tells the agent to run 'claude mcp add auxiliar -- npx auxiliar-mcp' and then call solve_task/list_solve_tasks. The instructions themselves do not ask for unrelated files or env vars, but they explicitly cause the agent to download and execute third‑party code and to surface candidate install commands (pip install, apt, etc.). Those candidate installs may request API keys or access to other services — the document warns of that but does not declare those credentials up front.

ℹ Install Mechanism

This is instruction-only (no coded install spec), but it directs runtime use of npx to fetch and run auxiliar-mcp from the npm ecosystem. npx (and downstream pip/apt install commands shown in candidate install strings) will download and execute remote code — a moderate supply‑chain risk. The URLs referenced (GitHub repos, homepage) look plausible, but no pinned release URLs or checksums are provided in the skill text.

ℹ Credentials

The skill declares no required env vars or credentials, which is consistent with its role as a ranking service. However, its recommendations may include cloud services or vendor APIs (e.g., Google Document AI) that require API keys; the agent may be instructed to request or use those credentials during follow-up operations. The skill does not proactively request broad unrelated credentials.

✓ Persistence & Privilege

always:false and the skill does not request system-wide configuration changes or modify other skills. Autonomous invocation is allowed (platform default) but not combined with 'always:true' or other elevated privileges.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install auxiliar-solve
安装完成后，直接呼叫该 Skill 的名称或使用 /auxiliar-solve 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

vv0.1.0

auxiliar-solve v0.1.0 - Initial release of the auxiliar-solve skill for agent task tool ranking. - Supports ranked recommendations for OCR, PDF extraction, Brazilian NFS-e invoices, boletos, receipts, web scraping, and bookkeeping. - Provides reproducible evals on real-world corpora, with detailed scorecards and installation instructions. - Enables agents to discover, compare, and select tools based on accuracy, speed, cost, and compatibility. - Offers transparent evaluation methodology, alternatives considered, and agent compatibility information.

元数据

Slug auxiliar-solve

版本 v0.1.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题

Auxiliar Solve 是什么？

Ranked installable tools for agent jobs — OCR, PDF extraction, NFS-e invoices, bookkeeping, boletos, receipts, web scraping. Reproducible evals on real-world... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 78 次。

如何安装 Auxiliar Solve？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install auxiliar-solve」即可一键安装，无需额外配置。

Auxiliar Solve 是免费的吗？

是的，Auxiliar Solve 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

Auxiliar Solve 支持哪些平台？

Auxiliar Solve 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。

谁开发了 Auxiliar Solve？

由 Tlalvarez（@tlalvarez）开发并维护，当前版本 vv0.1.0。