Swot Analysis Agent Skill

SWOT Analysis (Evidence-Cited)

Purpose

Build an evidence-cited SWOT of one company — yours or a competitor's — from the public record:

use or gather evidence → four labeled quadrants → crossings → next-step options. Strengths and

weaknesses come from documented capability and customer voice; opportunities and threats from market,

competitor, and macro signals. Every entry carries a source and a label, and the analysis ends with

the S-O and W-T crossings — because a SWOT without "so what" is a parking lot with four sections.

Input

Works best with: the company to analyze, the scope (whole company or one product line/market),

and the decision this SWOT should support.

Also useful: the perspective — theirs, or yours competing with them — and any research already in

session (a competitive-research-snapshot, VoC themes, a

landscape scan, company-intel output): the SWOT then *synthesizes*

instead of re-searching.

Input supplied inline with the invocation — text after the skill name, a pasted context dump, or an

appended ARGUMENTS: line — counts as answers already given. Use it against the question budget;

don't re-ask.

Arriving empty-handed? That works too. The skill opens with at most 3 questions (company + scope,

decision, perspective) and proceeds on labeled assumptions if they go unanswered.

Example invocation: `Evidence-cited SWOT on [Competitor A]'s mid-market product line, from our

perspective — supports the Q3 positioning refresh. Last month's snapshot is attached.`

Key Concepts

• Governing protocol: honors the autonomous-investigation

contract — question budget of 3, search-plan gate, Fact/Inference/Assumption labels, Just Enough

Mode (max 5 ranked entries per quadrant), stable schema, 4-option Final Step. Disciplines: OSINT +

FININT + HUMINT per intelligence-collection-disciplines.

• Quadrant discipline. S and W are *internal and current*; O and T are *external*. The most

common SWOT error is an "opportunity" that requires the company to act — that's a strategy, not an

opportunity. Flag it when you catch it; catching it is half the teaching.

• Claimed vs. evidenced advantage. Strengths sourced from the company's own marketing are

claims; strengths corroborated by customer voice, filings, or independent coverage are evidence.

Customer voice weighs heaviest on weaknesses — buyers document weaknesses vendors won't.

• The crossings are the deliverable. S-O names the attack (which strength unlocks which

opportunity); W-T names the exposure (which weakness a threat finds first). SWOT is an input to

moves, not a wall poster.

• Companion, not substitute: porters-five-forces reads the

*industry's* structure; this reads *one company's* position within it. Run both when the decision

is market entry.

• When NOT to use: internal-only SWOT from private knowledge → run it as a team workshop with

your evidence; the whole competitor set → snapshot first, SWOT the one or two that matter.

• Do-not-invent list: financials, customer wins, market share, product claims, roadmap items.

Application

• Check session for existing evidence (snapshot, VoC themes, landscape scan, company-intel).

Present → synthesize from it; search only gaps and staleness.

• Credit inline context, then ask only the unanswered questions (max 3):
• Which company, and whole-company or one product line?
• What decision should this support?
• From whose perspective — theirs, or yours competing with them?
• If researching fresh, show the 3-bullet search plan — source sweep (filings, reviews, press,

pricing pages, hiring signals), how internal vs. external evidence will be separated, how facts

split from inference. Continue unless revised.

• Emit the schema below exactly.

Output schema (do not reorder)

~~~markdown

SWOT: [Company / Scope]

As-of date: | Decision supported: | Perspective:

1. Strengths (internal, evidenced)

• [Entry] — [evidence: URL, date] — [Fact/Inference]
• [Max 5, ranked by defensibility]

2. Weaknesses (internal, evidenced)

• [Entry — customer voice weighs heaviest here] — [URL] — [label]
• [Max 5, ranked by exploitability]

3. Opportunities (external)

• [Entry — market/macro signal] — [URL] — [label]
• [Max 5, ranked by fit to their strengths]

4. Threats (external)

• [Entry — competitor/regulatory/technology signal] — [URL] — [label]
• [Max 5, ranked by likelihood x damage]

5. The Crossings (the "so what")

• S-O (attack): [strength that unlocks which opportunity — 1-2 moves, labeled Inference]
• W-T (exposure): [weakness a threat will find first — 1-2 risks, labeled Inference]
• For your decision: [what this SWOT says about the decision named above, in 2 sentences]

Assumptions to Validate

• [Assumption 1] / [Assumption 2] / [Assumption 3]

~~~

A copy/paste fill-in version of this schema, with quality checks, lives in template.md.

Final Step (offer exactly 4 options)

• Run the W-T exposure into a battle card or risk plan (battle-card-builder)
• SWOT a second company for side-by-side comparison
• Feed the crossings into a positioning statement (positioning-statement)
• Schedule-ready version: which entries should a quarterly re-run watch?

Accept 1, 2, 3, 4, 1 and 2, Verbose Mode, or a custom path.

Examples

The quadrant discipline catching a disguised strategy (fictional):

> Draft opportunity: "expand into the compliance-automation segment." Caught and reclassified — that

> requires *them* to act, so it's a candidate strategy, not an opportunity. The underlying

> opportunity, restated externally: "compliance-automation demand in their install base is growing —

> Fact ([industry survey, URL]) — and no incumbent serves it." Now the S-O crossing can test

> whether any *documented strength* (their services arm, their audit-trail architecture) actually

> unlocks it — or whether it's a mirage that flatters the roadmap.

Claimed vs. evidenced, side by side:

> - Their site claims "enterprise-grade reliability" — that's a claim, not a strength.

> - Strength as evidenced: "publishes a status page with 26 months of >99.95% uptime — Fact

> ([status page, URL]) — and reliability praise recurs across independent reviews — Fact

> ([review site, URL])." Ranked #2 by defensibility: verifiable, sustained, hard to copy quickly.

See examples/sample.md for a complete worked SWOT (fictional FSM-software

market, same scenario as the suite's chain examples) that synthesizes from a snapshot and watch

report already in session. examples/sample-industrial.md runs the

same quadrant discipline on an industrial evidence diet, including a fortress-move warning in the

S-O crossing.

Common Pitfalls

• The four-box brainstorm. Unsourced entries from a conference room are opinions arranged in

quadrants. Evidence per entry is what separates this SWOT from the one nobody trusts.

• Opportunities that are strategies. "Launch in Europe" is a move; "European demand outpacing

local supply" is an opportunity. If the entry needs the company to act, reclassify it.

• Marketing-sourced strengths. A strengths quadrant built from the company's own website is

their positioning statement wearing your logo. Corroborate or label as claim.

• Skipping the crossings. Four tidy quadrants and no moves. The S-O attack and W-T exposure are

why the analysis exists — the decision-support sentence at the end is the test it must pass.

• Wish-list contamination. Internal hopes ("we'd love better onboarding") placed in W, or

fears in T, without evidence. Every quadrant works the public record; wishes go to the roadmap.

References

• autonomous-investigation (Workflow) — the governing protocol
• intelligence-collection-disciplines (Component) — the evidence channels behind each quadrant
• porters-five-forces (Workflow) — industry structure; run both for market-entry decisions
• competitive-research-snapshot (Workflow) — upstream evidence base
• voice-of-customer-miner (Workflow) — the weaknesses quadrant's best source
• positioning-statement (Component) — consumes the crossings
• SWOT framework as standard strategy practice; adapted from

market-intelligence/swot-analysis-prompt.md in the

https://github.com/deanpeters/product-manager-prompts repo.