What makes agent-analytics-autoresearch notable?

1.2K downloads. Trust evidence available

How should agent-analytics-autoresearch be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

agent-analytics-autoresearch

Run an autoresearch-style growth loop for landing pages, onboarding, pricing, and experiment candidates. Collect or read analytics snapshots, preserve product truth, generate/critique/synthesize variants, blind-rank with Borda scoring, and output two review-ready A/B test variants. Works with any analytics data; best with Agent Analytics CLI/API. Skill: agent-analytics-autoresearch Owner: dannyshmueli Summary: Run an autoresearch-style growth loop for landing pages, onboarding, pricing, and experiment candidates. Collect or read analytics snapshots, preserve product truth, generate/critique/synthesize variants, blind-rank with Borda scoring, and output two review-ready A/B test variants. Works with any analytics data; best with Agent Analytics CLI/API. Tags:

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.2k

Updated

Oct 11, 2026

Version

1.0.12

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.2K downloads reported by the source. Last updated 10/11/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 11, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 11, 2026
Adoption signal: 1.2K downloadsadoption · observed Oct 11, 2026
Latest release: 1.0.12release · observed Oct 9, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s1785jgb7evz92f88z9psd4m7583gmsv:agent-analytics-autoresearch

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-dannyshmueli-agent-analytics-autoresearch/snapshot"

Documentation

CLAWHUB

147,282 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: agent-analytics-autoresearch
description: "Run an autoresearch-style growth loop for landing pages, onboarding, pricing, and experiment candidates. Collect or read analytics snapshots, preserve product truth, generate/critique/synthesize variants, blind-rank with Borda scoring, and output two review-ready A/B test variants. Works with any analytics data; best with Agent Analytics CLI/API."
version: 1.0.12
author: dannyshmueli
license: MIT
repository: https://github.com/Agent-Analytics/skills
homepage: https://agentanalytics.sh
compatibility: Requires a coding agent that can read and write local files. Agent Analytics data collection requires npx and browser approval or detached login. The loop can also run from pasted reports, SQL output, CSV exports, or existing analytics files.
tags:
- analytics
- autoresearch
- growth
- experiments
- ab-testing
- landing-pages
provides:
- capability: autoresearch
- capability: ab-testing
- capability: growth-experiments
- capability: landing-page-optimization
metadata:
openclaw:
requires:
anyBins:
- npx
---

# Agent Analytics Autoresearch

Use this skill when the user wants a data-informed growth loop for landing pages, onboarding, pricing, CTAs, signup, checkout, activation, or other experiment candidates.

This skill is based on:

- Autoresearch Growth template: <https://github.com/Agent-Analytics/autoresearch-growth>
- Agent Analytics: <https://agentanalytics.sh>
- Regular Agent Analytics skill: <https://github.com/Agent-Analytics/skills/tree/main/skills/agent-analytics>

Use the regular `agent-analytics` skill for general setup, tracking installation, ad hoc reporting, and normal experiment operations. Use this skill for structured variant generation and judging from a project brief plus analytics data.

## Core Rule

Do not edit production copy, product code, or live experiment setup while running the loop unless the user explicitly asks. Produce reviewable artifacts first.

Default mode is review-only: generate variants, log rounds, and write `final_variants.md`.

After explicit human approval, continue into the outer experiment loop when requested: implement the approved variant or variants, create the experiment, run it, measure it with Agent Analytics or another analytics source, save the results as the next snapshot, and start the next autoresearch run from evidence.

## Inputs

The loop needs:

- target surface
- current control copy
- product truth
- audience
- primary metric
- proxy metric
- guardrails
- analytics snapshot or data brief
- drift constraints

Agent Analytics is preferred, but not required. Accept any evidence source: Agent Analytics CLI/API, PostHog, GA4, Mixpanel, SQL, CSV exports, product logs, dashboard screenshots summarized by the user, or hand-written notes.

When Agent Analytics is the evidence source, use project context as the self-improving product memory for the loop. Read `context get <project>` before collecting a snapsh

_meta.json

{
  "ownerId": "kn7caxjvqk9fengp67p290smnn800sv9",
  "slug": "agent-analytics-autoresearch",
  "version": "1.0.12",
  "publishedAt": 1791547065276
}

references/brief-template.md

# Growth Loop Brief

## Target

- Project:
- Surface:
- Public URL:
- Local source file or copy source:
- Primary metric:
- Proxy metric:
- Guardrail metrics:
- Recommended experiment name:
- Variant shape:

```text
variants: control,candidate_1,candidate_2
goal: <primary_event>
proxy: <proxy_event>
```

## Product Truth

Describe what the product is, who it serves, and what must remain true in every candidate.

Include:

- core promise
- target audience
- strongest differentiator
- language the product should own
- claims the product can support
- claims the product should not make

## Audience

Primary audience:

-

Pain:

-

Desired action:

-

## Current Control

Headline:

```text

```

Subheadline:

```text

```

Primary CTA:

```text

```

Supporting copy:

```text

```

## Analytics Commands Or Data

List commands, API calls, SQL queries, exports, or pasted data the agent should use before generating variants.

Agent Analytics CLI example:

```bash
# Run once if this machine or agent runtime is not logged in.
npx --yes @agent-analytics/[email protected] login

PROJECT_SLUG=<project_slug>
PRIMARY_EVENT=<primary_event>
PROXY_EVENT=<proxy_event>
RUN_DATE=$(date +%F)

mkdir -p "data/$RUN_DATE"

# Keep collecting the full snapshot even if one analytics command fails.
# Failed commands write their error output and exit code into the saved file.
run_snapshot_command() {
  output_file="$1"
  shift
  set +e
  "$@" > "$output_file" 2>&1
  command_status=$?
  set -e
  perl -i -pe 's/\e\[[0-9;]*m//g' "$output_file"
  if [ "$command_status" -ne 0 ]; then
    printf '\ncommand_exit_code: %s\n' "$command_status" >> "$output_file"
  fi
}

run_snapshot_command "data/$RUN_DATE/insights.txt" npx --yes @agent-analytics/[email protected] insights "$PROJECT_SLUG" --period 7d
run_snapshot_command "data/$RUN_DATE/pages.txt" npx --yes @agent-analytics/[email protected] pages "$PROJECT_SLUG" --since 7d
run_snapshot_command "data/$RUN_DATE/funnel.txt" npx --yes @agent-analytics/[email protected] funnel "$PROJECT_SLUG" --steps "page_view,$PROXY_EVENT,$PRIMARY_EVENT" --since 7d
run_snapshot_command "data/$RUN_DATE/${PROXY_EVENT}-events.txt" npx --yes @agent-analytics/[email protected] events "$PROJECT_SLUG" --event "$PROXY_EVENT" --days 7 --limit 50
run_snapshot_command "data/$RUN_DATE/${PRIMARY_EVENT}-events.txt" npx --yes @agent-analytics/[email protected] events "$PROJECT_SLUG" --event "$PRIMARY_EVENT" --days 7 --limit 50
run_snapshot_command "data/$RUN_DATE/experiments.txt" npx --yes @agent-analytics/[email protected] experiments list "$PROJECT_SLUG"
```

Generic placeholders:

```bash
<analytics-command> summary --project <project> --period 7d
<analytics-command> pages --project <project> --period 7d
<analytics-command> events --project <project> --event <proxy_event> --period 7d
<analytics-command> events --project <project> --event <primary_event> --period 7d
<analytics-command> funnel --project <project> --steps "page_view,<proxy_event>,<primary_event>"
<analytics-command> experiments --project <project> li

references/final-variants-template.md

# Final Variants

## Target Experiment

```text
experiment:
variants: control,candidate_1,candidate_2
goal:
proxy:
```

The experiment has not been wired yet.

## Control

Headline:

```text

```

Subheadline:

```text

```

Primary CTA:

```text

```

Supporting copy:

```text

```

## Candidate 1

Headline:

```text

```

Subheadline:

```text

```

Primary CTA:

```text

```

Supporting copy:

```text

```

Hypothesis:

Risks:

## Candidate 2

Headline:

```text

```

Subheadline:

```text

```

Primary CTA:

```text

```

Supporting copy:

```text

```

Hypothesis:

Risks:

## Why These Two

## Data Notes

## Next Step

references/program.md

# Autoresearch Growth Loop

You are running an instruction-driven growth loop. Produce two high-quality variants that can be tested against the current control for the project described in `brief.md`.

Do not change product code while running the loop. Produce reviewable copy artifacts first.

Default mode is review-only. Only move into implementation, experiment creation, or measurement after the user explicitly approves that next phase.

## Setup

1. Read `brief.md` fully.
2. Treat `brief.md` as the source of truth for the project, audience, surface, control, metrics, analytics data, and drift constraints.
3. If `results.tsv` does not exist, create it with:

```tsv
round candidate_a candidate_b candidate_ab winner borda_a borda_b borda_ab status rationale
```

4. If `final_variants.md` exists, overwrite it only at the end of a completed loop.

## Data Brief

Before generating copy, collect or read the data described in `brief.md`.

The data source can be Agent Analytics, another analytics CLI, an API, SQL, CSV, exported reports, product logs, or manually supplied data.

Rules:

- Never invent numbers.
- If a command fails, record the failure.
- Treat missing data as missing data, not zero data.
- Separate primary outcome data from proxy and guardrail data.
- If data is sparse, say so and treat it as weak signal.

## Scope

During the loop, edit only:

- `results.tsv`
- `final_variants.md`
- optional scratch notes

Do not edit the live site, app, product code, or experiment setup until the variants have been reviewed.

## Product Truth

Every candidate must preserve the product truth from `brief.md`.

Penalize:

- generic category language
- copy a competitor could say word for word
- unsupported claims
- drift away from the real product value
- clickbait that weakens primary conversion intent
- changes that ignore the current control's strengths

Reward:

- specificity
- clear audience fit
- concrete user outcome
- stronger primary-event intent
- honest use of the available data
- language only this product could credibly say

## Loop

Run at least 5 rounds unless `brief.md` specifies a different count.

Each round has four phases.

### 1. Candidate A

For round 1, candidate A is your first new hypothesis based on the brief, control, and data.

For later rounds, candidate A is the previous round winner.

Include the editable parts named in the brief, usually headline, subheadline, CTA, supporting copy, and hypothesis.

### 2. Critique

Critique candidate A harshly:

- what is generic
- what a competitor could say
- where value is unclear
- where copy drifts from product truth
- whether primary-event intent is strong enough
- whether the control is clearer

### 3. Candidate B And Synthesis AB

Write candidate B from the critique as if starting fresh. It may use the brief and critique, but it should not be a mild rewrite of A.

Then write candidate AB by combining the strongest parts of A and B. Do not average them into bland middle copy.

### 4.

Github ReposUpdated 1d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.