Skip to content
Back to skills

Aca Experiment Design

ASecurity

Design one controlled ACA experiment for outbound or content. Use when the user wants to improve reply rate, test an angle, split audiences, A/B test copy, compare channels, or plan experiments.

  • 2 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 3, 2026
ai-agents

Works with

  • mcp

Security analysis

A100/100

Scanned September 3, 2026

npx -y skills add seguelaCedric/aca-native-mcp-skills --skill aca-experiment-design --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Aca Experiment Design?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Aca Experiment Design
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/seguelacedric-aca-experiment-design/badge)](https://www.skillsdirectory.com/skills/seguelacedric-aca-experiment-design)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: aca-experiment-design
description: "Design one controlled ACA experiment for outbound or content. Use when the user wants to improve reply rate, test an angle, split audiences, A/B test copy, compare channels, or plan experiments."
license: MIT
compatibility: Requires the ACA MCP server connected with a valid bearer token.
metadata:
  author: ACA
  version: "1.0"
  homepage: https://aca.so/product/integrations/mcp
---

# ACA experiment design

Create a clean experiment with one variable, one metric, and a stopping rule.

## Rules

- Do not change multiple variables at once.
- Prefer small tests before scaling.
- Save the experiment plan before applying changes.
- Mutations require approval.

## Workflow

1. Inspect current assets with `list_campaigns`, `list_email_sequences`, `list_lead_lists`, `list_generation_jobs`, and `list_autopilots`.
2. Pick experiment type:
   - Audience
   - Angle
   - CTA
   - Channel
   - Sequence timing
   - Content topic
3. Define control, variant, sample size, metric, decision threshold, and review date.
4. Save with `create_strategy_document`.
5. If approved, route to the right executor:
   - `aca-copy-variants`
   - `aca-find-leads`
   - `aca-launch-outreach`
   - `aca-content-week`

## Output format

```text
Experiment: {name}
Hypothesis: {hypothesis}
Control: {control}
Variant: {variant}
Metric: {metric}
Sample size: {n}
Decision rule: {rule}
Next action: {skill}
```

## Skill chaining

This skill participates in the ACA chain. Preserve the selected ACA org, relevant IDs, user brief, approval state, and any generated artifacts when continuing into another ACA skill. If the user asked for execution and a downstream condition is met, continue into the next skill automatically; otherwise end with the handoff block.

**Upstream**
- Called by `aca-weekly-rhythm`, `aca-positive-reply-scoring`, `aca-copy-variants`, or optimization requests.

**Auto-continue conditions**
- Copy variable -> continue to `aca-copy-variants`.
- Audience variable -> continue to `aca-find-leads` or `aca-lead-quality`.
- Content variable -> continue to `aca-content-week`.
- Launch-ready test -> continue to `aca-launch-outreach`.

**Stop before chaining when**
- Ask before applying experiment changes or launching tests.

**Downstream skills**
- `aca-copy-variants` - create message variants.
- `aca-find-leads` - build audience variant.
- `aca-lead-quality` - segment audience variant.
- `aca-content-week` - create content variant.
- `aca-launch-outreach` - launch approved experiment.

**Handoff block**

```text
Chain state: {continue|needs_approval|blocked|complete}
Next skill: {aca-skill-name|none}
Reason: {why this handoff is or is not needed}
Carry forward: {org_id/name, product_id, icp_id, lead_list_id, campaign_id, sequence_id, job_id, approvals, constraints}
```

## ACA tools used

- `list_campaigns`, `list_email_sequences`, `list_lead_lists`
- `list_generation_jobs`, `list_autopilots`
- `create_strategy_document`

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.