How to use the Strategy Red-Team agent skill (strategy-red-team)

What the strategy-red-team skill makes the agent do

The skill focuses on load-bearing assumptions: the claims that, if false, would make the plan fail. The agent lists what the plan asserts about the user, market, constraints, mechanism and timeline, then separates the critical claims from cosmetic ones.

For each load-bearing claim, the agent first states the strongest case for it, then attacks that version. Each failure mode is written as a concrete, testable 'Fails if' condition and ranked by impact × likelihood × cheapness to test, so the cheapest high-impact checks come first.

• Fails if: the precise condition that would break the plan
• Evidence to get this week: the specific data, query or conversation that would confirm or kill the assumption
• Kill criterion: the threshold at which to stop or change course
• Cheapest test: the smallest experiment that moves the belief
• What's Well-Reasoned: claims the skill judges sound, stated explicitly
• What I Couldn't Assess: gaps where the plan did not give enough to judge

Install the skill from the ZIP

Download the ZIP for the skill from the directory page, then unzip it. The source material for this skill is a single file, SKILL.md, inside the pm-execution/skills/strategy-red-team folder of the phuryn/pm-skills repository.

For a personal install, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder so the skill applies only there.

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

• Personal: ~/.claude/skills/
• Project only: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

Trigger the skill with natural requests

You do not need a special command. The skill's description lists the situations it applies to, so ask for what you want in ordinary words. Paste or attach your PRD, roadmap or strategy, then include a request such as the ones below.

The skill is also built to take the plan you provide as its input, so the more concrete the plan, the more useful the review.

• Stress-test this PRD before I take it to the exec review.
• Pressure-test this strategy and tell me which assumptions to check first.
• Challenge the assumptions in this roadmap.
• Help me prepare this doc for executive review.

Read and act on the output

The output opens with a ranked list of three to five top kill-assumptions. Start with the first item: it is the one the skill ranks as high-impact, plausibly wrong and cheap to check. Run its evidence check this week before you commit more resources.

Then read the sections on what holds up and what could not be assessed. A good red-team should say when a claim is sound rather than invent doubt.

• Work the ranked list from the top
• Use the kill criterion as a pre-agreed stop line
• Treat 'What I Couldn't Assess' as a list of information to add to the plan

Safety notes: scripts and network

According to the listing, this skill runs no scripts and uses no network access. It requires nothing beyond the plan you give it. The skill is text instructions only.

The agent cannot verify market facts on its own. It can reason about gaps in your plan, but it cannot confirm your data, so check every evidence claim against real customer or market sources.

• Runs scripts: no
• Uses network: no
• Requires: nothing listed
• Licence: MIT

Tips for better red-team results

The skill can only attack claims that are written down. Plans that are vague give it little to work with, and it will say so in the gaps section.

Ask for a second opinion only when you need one. The default is a single model, and the skill describes an optional cross-model mode where a second model such as Codex, Gemini or another Claude is reachable and flags where the two disagree.

• Include the user, market, constraint, mechanism and timeline in the plan
• State your numbers and assumptions explicitly
• Use the output to decide what to test, not as proof the plan is right