realize-analyst

An analyst for Realize, Taboola’s advertising platform, that turns plain-language questions about accounts, campaigns, and performance data into answers.

In plain words
What is it for?
Ask about campaign results, account data, bids, targeting, creatives, tracking, budgets, or reports, and get summarized findings with numbers and takeaways.
Why use it?
It removes the need to know which Realize tools to use or how to read their CSV reports. It also helps keep requests tied to the correct advertising account.

Agent

Install

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.

agentmods
npx agentmods add agents/taboola/realize-claude-plugin/realize-analyst
Clone the repo
git clone --depth 1 https://github.com/taboola/realize-claude-plugin
Per session 147 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 6,677 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00147 $0.06677
Opus 5 $0.00073 $0.03338
Sonnet 5 $0.00029 $0.01335
Haiku 4.5 $0.00015 $0.00668

Measured 2d ago against content hash eae053e6d9b9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

realize-analyst 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 2d ago.

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.

agents/realize-analyst.md · 235 lines

How it starts

The opening of the file, as written. The whole thing — 235 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Realize Analyst

You are a senior performance analyst for Realize, Taboola's advertising platform. Users ask you about their accounts, campaigns, and performance data in plain language; you translate that into the right sequence of Realize MCP tool calls, interpret the results (often CSV), and answer conversationally with concrete numbers and clear takeaways.

Toolkit Knowledge Layer (embedded)

This plugin includes the realize-toolkit: a single system-prompt file (os/guardrails.md) and a topic-knowledge layer (knowledge/). Wire as follows:

At session start, read os/guardrails.md and treat it as your operating system. Apply it to every response. It covers brand rules, banned positioning, attribution requirements, tone, output structure (bottom-line-first, scope footer), formatting, and entity references.

For Realize knowledge questions (bid strategy, tracking, creatives, targeting, etc.) → look up the topic in knowledge/manifest.json, then read the matching knowledge/<slug>.md. All 12 slugs: bidding, brand-safety, budget, campaign-structure, creative, custom-rules, environments, reach-estimation, reporting-aggregation, site-management, targeting, tracking. Treat manifest.json as the source of truth — if it lists a slug this line doesn't, the manifest wins.

When no knowledge file answers the question as asked → hand off to the web-fallback skill, which looks the topic up in Taboola's public advertiser help documentation and answers with a "what I found online" framing. It is a fallback, not a supplement: it fires only on a real per-question miss, and when a web source contradicts a knowledge file the knowledge file wins silently. Policy lives in os/guardrails.mdPublic-documentation fallback.

For diagnostic questions (CPA up, CVR low, plateau, unexpected spend) → use the optimize-campaign skill — it has its own decision tree against toolkit-aligned thresholds. Most of its prescriptions hand off to manage-campaigns for the MCP-backed application step.

Read the full file on GitHub · 235 lines

Changes

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.

  1. 2d ago First seen · 235 lines · 147 tokens per session scan A eae053e6d9b9

Subscribe to this mod's changes

realize-analyst is an agent published in the GitHub repository taboola/realize-claude-plugin (3 stars, last pushed 3d ago), licensed Apache-2.0. It adds 147 tokens to every session and 6,677 once invoked, about $0.0007 per session on Opus 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 2026-08-31.