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.
npx skills add ChrisGVE/localdata-mcp --skill ab-testgit clone --depth 1 https://github.com/ChrisGVE/localdata-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/chrisgve/localdata-mcp/ab-test)<a href="https://agentmods.dev/skills/chrisgve/localdata-mcp/ab-test"><img src="https://agentmods.dev/badge/skills/chrisgve/localdata-mcp/ab-test/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/chrisgve/localdata-mcp/ab-test"><img src="https://agentmods.dev/badge/skills/chrisgve/localdata-mcp/ab-test.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00027 | $0.00711 |
| Opus 5 | $0.00014 | $0.00356 |
| Sonnet 5 | $0.00005 | $0.00142 |
| Haiku 4.5 | $0.00003 | $0.00071 |
Grade A, and why
ab-test scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 44 lines — stays where its author put it; the contents beside it link to each section on GitHub.
A/B Test Analysis
Analyze experiment results and deliver a clear ship, iterate, or no-ship recommendation.
Steps
-
Explore experiment data. Call
describe_databasewith the database name from$ARGUMENTS. Identify the table containing experiment results. Look for columns indicating: group assignment (treatment/control), the primary metric, and any segmentation variables. -
Inspect the data. Call
execute_queryto sample rows from the experiment table. Confirm the group labels (e.g., "control" vs "treatment", "A" vs "B"). Note the metric column name and type. Check for any timestamp or user ID columns. -
Check group balance. Call
execute_querywith a query that counts observations per group and computes mean/stddev of the primary metric per group. Verify that group sizes are roughly balanced (within 10% of each other). Flag any imbalance as a potential validity concern. -
Run the A/B test. Call
analyze_ab_testwith the database name, metric column, and group column. Review the test results: test statistic, p-value, confidence interval for the difference, and the detected difference between groups. -
Calculate effect sizes. Call
analyze_effect_sizeswith the same data. Get Cohen's d or the appropriate effect size measure. Classify the effect:- Small: d around 0.2
- Medium: d around 0.5
- Large: d above 0.8
-
Assess statistical significance. Determine if the p-value is below the significance threshold (typically 0.05). Note the confidence interval for the treatment effect. A significant result with a confidence interval that excludes zero provides strong evidence.
-
Assess practical significance. Compare the effect size against business-relevant thresholds. A statistically significant result with a tiny effect size may not justify shipping. Conversely, a marginally insignificant result with a meaningful effect size may warrant further testing.
-
Make a recommendation. Apply this decision framework:
- Ship: p-value below 0.05 AND effect size is practically meaningful AND confidence interval is entirely positive (or negative, depending on desired direction)
- Iterate: effect is in the right direction but either not significant (need more data) or effect size is borderline (need refinement)
- No-ship: effect is zero, negative, or opposite to the desired direction with sufficient statistical power
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.
- 10d ago First seen · 44 lines · 27 tokens per session scan A 82c36628140e
ab-test is a skill published in the GitHub repository ChrisGVE/localdata-mcp (4 stars, last pushed 26d ago), licensed Apache-2.0. It adds 27 tokens to every session and 711 once invoked, about $0.0001 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.
Other skills, from other repositories
wingman-mcp
Add persistent, interactive task plan management to your agent. Wingman tracks plans and tasks across a long conversation — Claude creates plans, ticks tasks after completing work, and a live panel renders inline in the chat. Install the MCP server first, then use these tools to manage plans and tasks throughout any…
liveagent-code-review
Review an open GitHub pull request or the current local branch and working tree with parallel, independent reviewers and evidence-based validation. Use when the user asks for code review, invokes the Code Review action from Git Review, or explicitly mentions this skill.
skills-creator
Create or update LiveAgent runtime skills. Use when you need to summarize a workflow into a skill, write it into the fixed user skills root, validate it, package it, or refine a skill's references for LiveAgent's SkillsManager flow.
skills-installer
Install, list enabled, validate, or package LiveAgent skills. Use when you need to inspect the skills enabled in the current conversation, import a local skill directory or package, search/install from ClawHub, install from a GitHub repo/tree URL, or reconcile conflicts during an upgrade.
triaging-incidents
Creating, triaging, updating, and analyzing ServiceNow incidents. Assess severity, check affected CIs, find related incidents, and recommend assignments. Use when the user mentions incidents, INC numbers, outages, service disruptions, ticket creation, triage, priority, severity, ITSM operations, SLA breaches…
reviewing-update-sets
Reviewing, analyzing, and comparing ServiceNow update sets before promotion. Shows changes, risks, dependencies, and conflicts. Use when the user mentions update sets, customizations, promotion, code review, change tracking, pre-deployment review, sysupdatexml, customer updates, "what changed in this update set," or…