> ## Documentation Index
> Fetch the complete documentation index at: https://developers.everflow.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Agent Skills
> Packaged, auto-activating workflows that run common partner-marketing jobs reliably and token-efficiently on top of the Everflow MCP tools.
**Agent Skills** package a complete workflow — the procedure, the judgment, and the exact tool calls — so an AI agent runs it reliably instead of re-deriving it each time. They sit on top of the MCP [tools](/ai-automation/mcp/tools): the tools are the raw verbs; a skill is the playbook that strings them together and knows how to interpret the result.
**Skills vs. the [Prompt Library](/ai-automation/mcp/prompts).** Freeform prompts are text you paste in. Skills are **auto-activating** — the agent loads the right one when your request matches it (no copy-paste, no remembering tool names) — and use **progressive disclosure**, so only a short name + description is in context until the skill is actually needed (token-efficient). Skills are the recommended way to get repeatable, high-quality results.
## Available skills
| Skill | Type | What it does |
| ------------------------- | ---- | ------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **performance-review** | read | Period performance review & pacing — headline numbers, period-over-period movement, top movers, with interpretation (not a data dump) |
| **tracking-setup** | read | Retrieve & verify a partner's tracking link for an offer — run-gate authority + the plumbing (domains, destination URLs, the `{transaction_id}` macro) |
| **conversion-scrubbing** | read | Review conversions and recommend which to approve/reject (scrub) — with evidence |
| **traffic-quality-audit** | read | Find pockets of problematic traffic — invalid/fraud/low-quality sources, ranked by *rate* and named by error code |
| **funnel-integrity** | read | Find where a conversion funnel leaks or a pixel misfires, using the offer's advertiser-event goals |
| **yield-optimizer** | read | Find underspending winners — efficient offers/partners with real signal and cap headroom to scale |
| **geo-optimizer** | read | Rank geos to scale or cut on volume-qualified evidence, with a suggested reallocation |
The skills work on any Network key with the matching read permission.
Every skill above is published to the one-command discovery endpoint. You can also [install any of them manually](#installing-the-skills).
## Installing the skills
The **read** skills are published straight from this documentation site — Everflow serves the standard [Agent Skills](https://agentskills.io) discovery manifests automatically at `https://developers.everflow.io/.well-known/agent-skills/index.json` (0.2.0 spec, with integrity digests) and `…/.well-known/skills/index.json` (legacy), each `SKILL.md` at a stable URL. Add them with one command:
```bash theme={null}
npx skills add https://developers.everflow.io
```
This works with any Agent Skills–compatible client (Claude Code, Cursor, Gemini CLI, and others). To install just one, append `--skill ` (e.g. `--skill tracking-setup`).
You can also **install manually** by copying a skill's folder (`SKILL.md` + its `references/`) into your agent's skills directory; for Claude Code that's `~/.claude/skills/`.
Once installed, just describe what you want in plain language ("how did we do last week?", "get the tracking link for MediaBuy on ExpressVPN", "review pending conversions on offer 12") and the matching skill activates automatically.
The raw read tools each skill is built on.
The same workflows shown as step-by-step tool traces.
Freeform prompts when you'd rather drive the tools directly.