How to use the Business Assumption Stress Test agent skill

What the stress-test skill makes the agent do

The skill is part of the c-level-advisor/executive-mentor collection in the alirezarezvani/claude-skills repository. Its command is /em:stress-test followed by an assumption. It is meant for use before you commit to a plan whose core assumptions have not been validated, such as a claim that enterprise buyers will tolerate a 6-month pilot, or a hockey-stick revenue model.

When you ask the agent to stress-test an assumption, it follows a fixed method. It states the assumption as a specific, testable claim. It searches for counter-evidence, models downside cases such as a 30%, 50% or 80% shortfall, calculates how sensitive the outcome is to that one variable, and proposes hedges: a validation test, a fallback plan, or an early warning indicator.

The skill also includes question sets for revenue projections, market size, competitive moat, hiring plans and competitive response. It is a method and checklist written in a single file. It does not contain scripts or external data.

Install the stress-test skill

Download the ZIP for this skill from this page. The skill is distributed as a folder containing SKILL.md. Unzip it, then place the folder in one of two locations. For a personal install available across your projects, use ~/.claude/skills/. For a single project, use the .claude/skills/ folder inside that project.

If you use the Claude apps instead, you can upload the ZIP under Settings → Capabilities → Skills.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into the project's .claude/skills/ folder
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Trigger the skill with a request

The skill is invoked with the command /em:stress-test and an assumption written as a specific statement. The more precise the assumption, the more testable it is. Vague claims like 'our market is large' are hard to falsify, so the method works best when you give a number, a segment and a timeframe.

You can also describe the request in plain language. The skill's description covers pre-commitment reviews of unvalidated plans, so asking the agent to challenge a plan before you spend on it should bring it into play.

• Command form: /em:stress-test enterprise buyers will tolerate a 6-month pilot
• Plain request: Stress-test the assumption that our model reaches $2M ARR by December.
• Plain request: Challenge the market size in our investor deck before the board meeting.

Examples of assumptions to test

The skill's own examples show the kind of input it handles well. A useful test assumption names what is being claimed, who it applies to and what would have to be true.

Good inputs include a revenue model, a hiring timeline, a market-size claim, a retention figure, or a competitive scenario. Each one can be modeled with a downside case and a sensitivity check.

• Revenue: a new sales rep reaches full ramp in 2 months rather than 4
• Market size: the serviceable market is the number of named target accounts multiplied by average contract value
• Hiring: the plan still works if only 70% of planned headcount is hired
• Moat: churned customers would not have been kept by a competitor

What the output looks like

The skill defines a fixed output format. It starts by restating the assumption and its source, such as a model, an investor pitch or team gut feel. It then lists counter-evidence, a downside model at -30%, -50% and -80%, a sensitivity rating of high, medium or low, and a hedge section with validation, contingency and early warning items.

The downside model asks one key question at each level: does the business survive, and does the plan still make sense? Treat the result as a structured set of questions and a starting point for your own analysis, not as a finished forecast.

• Assumption and source
• Counter-evidence
• Downside model (bear, stress, catastrophic)
• Sensitivity rating
• Hedge: validation, contingency, early warning

Safety notes: scripts, network and data

The skill runs no scripts and makes no network calls. It needs no listed requirements. Its files are a single SKILL.md, so there is nothing to execute on install.

The skill instructs the agent to search for counter-evidence, but the files do not connect it to any data source. Whatever the agent finds depends on what it already knows or what you supply. Do not assume the output reflects verified market data.

• Runs scripts: no
• Uses network: no
• Requires: nothing listed
• Files: SKILL.md

Tips for better results

Supply the real numbers yourself. The output is only as good as the inputs you provide, so paste in your historical win rate, churn figures or hiring timeline rather than hoped-for values.

Test one assumption per run. The method measures sensitivity to a single variable, which gets muddied when several change at once. Work through the riskiest assumptions in your plan one by one.

• Use historical data, not targets, for win rates and churn
• Run one assumption per request
• Ask for the hedge section explicitly if you want concrete validation experiments