A Claude skill for analyzing a topic's background (how it came to be) and future (where it might go), using eight structured analytic frames drawn from intelligence tradecraft.
Built for people who need analysis with actual depth — researchers, analysts, journalists, strategists — and who are tired of watching AI produce five-bullet summaries in place of thinking.
You bring a topic. The skill picks the right frames for the question, states its selection out loud so you can override, then runs the frames to depth with required structural output — not free-form bullets.
The default is two frames, deep, not five frames, shallow.
Each frame comes with a built-in steel man (the strongest counter-reading to what was just produced) and a source quality note (weak-source claims, unverified assumptions, currency risk).
Backward-looking — how did we get here
- Genealogy — critical junctures that locked in today's configuration
- Actor Mapping — who wants what, stated vs revealed interests, coalitions
- Red Team — the situation through the analyzed actor's eyes
- Structural (STEEP) — social / technological / economic / environmental / political pressures
- Key Assumptions Check — what must be true for the current reading to hold
Forward-looking — where might it go
- Scenario 2x2 — plausible futures with observable signposts
- Second-Order Effects — consequences of consequences, out to third order
- ACH / Pre-Mortem / What-If — competing-hypothesis test, failure walkback, or shock scenario
Frames combine. The skill picks pairs by question type — an actor's move gets Actor Mapping + Red Team; a systemic shift gets Genealogy + Scenario 2x2; a contested reading gets ACH + Key Assumptions Check.
- "Why is X moving now?"
- "How did we get to this configuration?"
- "What might happen if Y decides Z?"
- "Whose reading of this holds up?"
- "What's the analysis missing?"
- Quick factual lookups
- Generic SWOT-style strategy templates
- Opinion pieces / hot takes
- Anything you already know the answer to
- Download the latest release ZIP from the Releases page.
- In claude.ai, go to Settings → Capabilities → Skills (or Customize → Skills, depending on your client version).
- Click Upload skill and select the ZIP.
Requires a paid Claude plan.
Clone the repo and symlink the skill folder:
git clone https://github.com/bushrabeg/structured-analysis.git ~/skills/structured-analysis
ln -s ~/skills/structured-analysis ~/.claude/skills/structured-analysisTo update: git pull inside the cloned repo. The symlink picks up changes.
The skill triggers automatically when your message signals background or future analysis of a topic. You don't need to invoke it by name.
Explicit frame:
"Run a red team on Ankara's Syria policy shift."
Trigger phrase:
"What's behind the recent uranium enrichment push in Iran?" → Skill picks Genealogy (triggered by "what's behind"), states its selection, invites additions, proceeds.
No signal — default match:
"Analyze the semiconductor decoupling." → Skill reads the topic as a systemic shift, picks Genealogy + Scenario 2x2, states its selection, proceeds.
Override count or direction:
"Just the background — three angles." → Skill runs three backward-looking frames.
Override the reflex:
"Skip the steel man on this one." → Steel man omitted; source note stays.
Before analysis begins, the skill outputs one line:
For this topic I've selected: Genealogy + Scenario 2x2. Add another or swap — otherwise I'm starting.
This is a transparency beat, not a permission request. Silence or "go" means proceed. An addition or swap redirects.
The bilingual trigger dictionary in SKILL.md has two columns — English and Turkish. Add your own by appending a column with the same rows. The skill will match user phrases in your language to the right frame.
Pull requests welcome for additional languages.
The eight frames draw from Structured Analytic Techniques (SATs) developed in the U.S. intelligence community and codified in the Kent School tradition — particularly Richards Heuer's Psychology of Intelligence Analysis and Structured Analytic Techniques for Intelligence Analysis (Heuer & Pherson).
Analysis of Competing Hypotheses, Key Assumptions Check, Red Team, and Pre-Mortem come directly from that tradition. Genealogy borrows from Foucauldian historical method. STEEP is standard futures-studies scaffolding. Scenario 2x2 is Shell / Global Business Network. Second-order effects and What-If are from strategic foresight practice.
This skill's contribution is not the frames themselves — it's the packaging: enforcing depth via required output sections, defaulting to a paired-frame minimum, and making frame selection transparent so the user stays in control.
Issues and pull requests welcome, especially:
- Additional language trigger dictionaries
- Topic types the default-match table doesn't cover
- Failure modes worth adding to the anti-patterns list
- Sharper phrasing anywhere the current text feels bureaucratic
Please keep the two-frame default and the transparency beat — those are load-bearing design choices, not decoration.
MIT. Use it, fork it, adapt it, ship your own version.