cto-red-team-scorecard

CTO Red-Team Scorecard

Overview

Use this skill to stress-test an agentic CTO against a set of real-world leadership challenges. The goal is not to praise the operating manual or repeat its principles. The goal is to expose where the CTO would likely fail in practice: missing data, weak delegation, incomplete change management, poor stakeholder handling, or vague execution artifacts.

The output should be a concise audit artifact that separates coverage from execution risk:

• what the docs already cover
• what would still fail in practice
• what artifact or behavior is missing
• how severe the gap is
• whether the challenge is a pass, partial pass, or fail

This is especially useful for:

• new-CTO challenge lists from blogs, interviews, or leadership playbooks
• internal stress tests of AGENTS.md / SOUL.md style operating manuals
• comparing an “answering” CTO doc against the messy realities of execution

When to Use

Use when:

• the user asks whether an agentic CTO would survive a set of external CTO challenges
• you need to turn a leadership article into a structured critique
• you want to identify practical gaps in strategy docs, not rewrite them
• the user asks for a red-team audit, stress test, or failure-mode analysis of CTO guidance

Do not use for:

• ordinary CTO advice with no evaluation component
• generic roadmap planning
• building operational templates unless the templates are explicitly the requested output

Workflow

1. Gather the challenge set.
   • Prefer a named source article if provided by the user.
   • If the user says “or any others you find online,” collect a small, diverse sample of similar external sources.
   • Ignore vendor pitches, product comparisons, and solution sales pages.
   • If the source is a PDF and direct download / remote navigation fails, try a local copy (e.g. file:///...pdf) in the browser viewer and extract page-by-page with thumbnails/page navigation.
   • When summarizing a source-backed recommendation, include the source explicitly in the output (for example, "Reference: McKinsey, ...") so the comparison is auditable.
   • If a PDF viewer renders but text extraction is empty, use visual page reads (browser vision) and page thumbnails instead of assuming the doc is unreadable.