/ecosystem-tool-evaluation

name: ecosystem-tool-evaluation
description: Evaluate and install ecosystem tools (Hermes plugins, skills, integrations) — assess complexity vs. benefit before installation, show examples for visual tools, and avoid "cloned but not configured" states.
version: 1.0.0
author: Nous Research
license: MIT
metadata:
  hermes:
    tags: [ecosystem, tools, installation, evaluation, user-preference]
    related_skills: [hermes-agent, spike]
origin: original
source_repo: kevinnft/ai-agent-skills
source_url: https://github.com/kevinnft/ai-agent-skills
source_license: MIT
language: en

Ecosystem Tool Evaluation

When users ask to "install all useful tools" or explore an ecosystem, evaluate each tool's complexity vs. benefit BEFORE installing. Avoid leaving tools in "cloned but not configured" states.

Core Principle

**Evaluate → Decide → Install (or Skip)**

Don't clone/install first and explain complexity later. The user's time and disk space are valuable.

Evaluation Framework

For each tool, assess:

1. **Setup Complexity**

| Level | Description | Examples | Action | |-------|-------------|----------|--------| | **Low** | Single command, no dependencies | `npm install -g tool`, `pip install tool` | ✅ Install immediately | | **Medium** | 2-3 commands, common dependencies | Install app + config file | ✅ Install with brief explanation | | **High** | Multiple steps, external services | Docker + database + API keys + integration | ⚠️ Explain complexity, ask before proceeding | | **Very High** | Infrastructure setup, ongoing maintenance | Self-hosted services, multi-component systems | ❌ Explain why it's complex, offer alternatives |

2. **Immediate Usability**

• **Ready after install?** (✅ tokscale, draw.io) → Install
• **Needs configuration?** (⚠️ API keys, config files) → Install + configure
• **Needs integration?** (❌ hindsight → Hermes) → Explain complexity first

3. **Value vs. Effort**

| Scenario | Setup Time | Value | Decision | |----------|-----------|-------|----------| | High value, low effort | 2-5 min | High | ✅ Install | | High value, medium effort | 10-15 min | High | ✅ Install with explanation | | High value, high effort | 30-60 min | High | ⚠️ Ask user first | | Low value, any effort | Any | Low | ❌ Skip or explain why not worth it | | Marginal value, high effort | 30+ min | Marginal | ❌ Skip, explain alternative |

User Preference Patterns

For "Gas" Users (Full Automation)

**User signal:** "gas dan otomatis aktif maksimal dan perfect lalu test pastikan sempurna"

**Pattern:** User wants FULL AUTOMATION with verification, not step-by-step confirmation.

**Workflow:**

1. Clone all repos
2. Install all skills
3. Configure automatically
4. Test & verify
5. Commit to git
6. Report results

NO intermediate confirmations.
NO "want me to proceed?" questions.
JUST DO IT and show final status.

**Example:**

❌ Bad:
"I found 4 trending skills repos. Want me to clone them?"
[waits for confirmation]
"Cloned. Want me to install?"
[waits for confirmation]

✅ Good:
"Gas! Cloning, installing, testing..."
[does everything]
"✅ COMPLETE: 64 skills installed, tested, committed (cdd7855)"

**When to use:**

• User says "gas" (go fast)
• User says "otomatis" (automatic)
• User says "maksimal" (maximize)
• User says "pastikan sempurna" (ensure perfect)

**What to do:** 1. Execute full workflow without asking 2. Test & verify at each step 3. Report final status with evidence 4. Only stop if critical error

**What NOT to do:**

• Don't ask "want me to proceed?"
• Don't wait for confirmation between steps
• Don't explain what you're about to do
• Just DO IT and report results

For Visual/UI Tools

**Don't ask "want to install X?" without context.**

**Do show examples first:**

❌ Bad:
"Want me to install draw.io for diagram generation?"

✅ Good:
"draw.io can generate diagrams like this: [show 3-4 example images]
These are useful for architecture diagrams, flowcharts, network topology.
Want me to install it? (2 minutes setup)"

**Why:** Users can't evaluate visual tools without seeing output. Show, don't tell.

For Complex Infrastructure

**Don't clone first, explain later.**

**Do evaluate complexity upfront:**

❌ Bad:
[clones hindsight]
"hindsight is cloned but not configured because it needs Docker + PostgreSQL..."

✅ Good:
"hindsight requires:
- Docker + PostgreSQL (30 min setup)
- LLM API key (additional costs)
- Hermes integration code
- Ongoing maintenance (server 24/7)

Current built-in memory is sufficient for single-user use.
Want to proceed anyway, or skip?"

**Why:** Avoid "cloned but not configured" states. Either install fully or don't install.

For "Install All Useful" Requests

**Don't install everything blindly.**

**Do categorize by readiness:**

1. **Install immediately:** Low complexity, high value (tokscale, skills libraries) 2. **Install with explanation:** Medium complexity, high value (draw.io + xvfb) 3. **Explain and ask:** High complexity, high value (web dashboards, multi-agent) 4. **Skip and explain:** High complexity, marginal value (hindsight for single user)

For Tool/Project Analysis Requests

**When user asks to analyze a tool/project, assess relevance FIRST before dumping features.**

**Pattern:** 1. **Immediate relevance check** — "Is this urgent/important for current workflow?" 2. **If not urgent** — Say so directly, then provide analysis 3. **If urgent** — Lead with why it matters, then details

**Example:**

❌ Bad:
User: "Analisa Obsidian"
Agent: [dumps 12KB feature list immediately]

✅ Good:
User: "Analisa Obsidian"
Agent: "Gak ada yang urgent untuk saat ini. Semua skills itu optional — buat upgrade workflow lo aja kalau mau.

Yang paling immediately useful kalau lo sering:
1. Debugging masalah → systematic-debugging
2. Scraping/automation → web-scraping
3. Bikin custom skills → hermes-agent-skill-authoring

[then full analysis if user wants details]"

**Why:** User time is valuabl