strategic-brief

Strategic Brief Skill

This skill produces objective, evidence-based strategic analysis documents. It is designed to challenge assumptions, not confirm them. The output should be useful whether the recommendation is "proceed," "pivot," or "kill."

Before You Start

1. Read the analytical framework: references/framework.md in this skill's directory. It defines the 9-step analysis process, the document structures, and the quality gates.
2. Read the project: README, docs, recent git log, pyproject.toml / package.json, test suite, CI config. Understand what was built and why before evaluating it.
3. Identify the decision context: What triggered this analysis? A go/no-go decision? Architecture pivot? Budget review? The context shapes emphasis.

The Non-Negotiable Rules

1. No advocacy. You are an analyst, not a salesperson. If the evidence says "don't use this tool," say so clearly. Never soften a negative finding to protect the project.
2. Multiple reasoning paths. For every major conclusion, show at least two ways you arrived there. Name the reasoning path ("cost-driven analysis," "reliability-first analysis") and explain why you chose it.
3. Challenge assumptions explicitly. Dedicate a section to listing assumptions found in the code/docs and stress-testing each one. Some will hold. Some won't. Show your work.
4. Specific numbers over vibes. $/scan, $/year, rule counts, latency in seconds, pricing tiers. Vague comparisons ("much cheaper") are worthless in an executive brief.
5. Include "what would change this." Every recommendation has conditions under which it becomes wrong. Name them. This builds trust and gives the reader a framework for re-evaluating as conditions change.
6. Sources or it didn't happen. Every factual claim about a competitor, pricing model, or market trend needs a source. Group sources in an appendix. 20+ sources minimum.
7. 120-character line wrap. All generated markdown hard-wraps at 120 characters.

Research Phase

Before writing, complete these research tracks (use Explore agents in parallel):

Track 1: Internal Analysis

• What does the tool actually do? (Read source code, not just docs)
• How complex is the implementation? Could it be replaced?
• What's the test coverage and code quality?
• What assumptions are baked into the architecture?
• What's the maintenance burden?

Track 2: Competitive Landscape

Research in concentric rings:

1. First-party/native: Does the platform vendor offer this capability?
2. Commercial tools: Who charges money to solve this problem?
3. Open-source: Free alternatives, community projects, framework plugins
4. Adjacent tools: CSPM, observability, governance platforms that partially overlap
5. Agent/LLM approaches: Could an AI agent do this without custom code?

For each competitor: what it does, pricing, rule/check count, deployment model, limitations.

Track 3: Architecture Alternatives

• Deterministic code vs LLM agent vs hybrid: cost, reliability, portability
• Build vs buy economics at 1x, 10x, 100x scale
• Security and regulatory implications of each approach
• What the market consensus pattern looks like

Document Production

Produce two documents. See references/framework.md for detailed structure templates.

Document 1: Full Strategic Analysis

Target audience: engineering leads, security team, technical stakeholders. Length: 2500-3500 words + appendices.

Sections (in order):

1. Executive Summary (1 page, board-readable even in isolation)
2. Problem Statement & Market Gap
3. Competitive Landscape (matrix with 15+ entries)
4. Build vs Buy Analysis
5. Architecture Decision: Agent vs Deterministic vs Hybrid
6. Assumptions Challenged
7. Recommendation & Decision Framework
8. Risk Register
9. Appendices (competitive matrix, cost models, sources)

Document 2: Executive Summary

Target audience: executive board, non-technical stakeholders. Length: 500-800 words. Must fit on 2 printed pages.

Sections (in order):

1. The Problem (2-3 sentences)
2. What We Built / What Exists (2-3 sentences)
3. Market Position (condensed comparison)
4. Key Architecture Decision (3 bullets max)
5. Recommendation (1 paragraph)
6. Next Steps (3-5 bullets)
7. Risk Summary (3 bullets)

Quality Gate

Before delivering, verify:

• Every major conclusion has 2+ reasoning paths shown
• At least 3 assumptions are explicitly challenged with counter-arguments
• Competitive matrix has 15+ entries with specific pricing/feature data
• Cost models use specific $/unit numbers, not relative comparisons
• "What would change this" section exists with concrete conditions
• Sources section has 20+ references
• No lines exceed 120 characters
• Executive summary works standalone (doesn't reference "see Section 5")
• Recommendation is clear and actionable, not hedged into uselessness