swot-batch.md

swot-batch.md

SWOT Output Format (every company, every level)

## SWOT: [Company]
**As-of date:** | **Evidence base:** [snapshot / fresh search]

### Strengths (internal, evidenced)
- [Entry] — [URL, date] — [Fact/Inference]  (max 5, ranked)

### Weaknesses (internal, customer voice weighs heaviest)
- [Entry] — [URL, date] — [label]  (max 5, ranked)

### Opportunities (external)
- [Entry] — [URL, date] — [label]  (max 5, ranked)

### Threats (external)
- [Entry] — [URL, date] — [label]  (max 5, ranked)

### Crossings
- **S-O (their likely attack):** [1-2 moves, Inference]
- **W-T (their exposure):** [1-2 risks, Inference]

Per-Company Receipt (closes each company)

#### Receipt: [Company]
- Entries: S[n] W[n] O[n] T[n] | Sourced: [n of total]
- Evidence quality: [strong / mixed / thin — why]
- Assumptions worth flagging: [list, or "none"]
- Status: [complete / skipped: reason / FAILED at: step]

Level 0 — Unseasoned (works today)

/batch Produce a SWOT analysis for each of these five competitors, with sources.

You'll get five SWOTs. Whether company four's weaknesses are evidence or improv is a question the output format won't answer.

Level 1 — Loop Lingo (the batch contract)

/goal Produce an evidence-cited SWOT for each competitor in my set —
every entry labeled and sourced per the SWOT Output Format, every
company closed with a receipt — then a cross-company comparison
built only from the settled SWOTs.

Base camp (calculate once): my product and segment, the decision
this supports, and the evidence already in session (snapshot,
landscape scan, VoC themes — reuse it; search only gaps).

Index before you batch:
1. List the companies (cap: 5 per run; more means run two).
2. Note which have session evidence vs. need fresh search.
3. Sequence: session-evidenced companies first (cheap before
   expensive).

/batch Process the list with these controls:

- One company at a time. Complete its SWOT and receipt before
  touching the next.
- Cheap checks first: is there enough public footprint to source a
  SWOT? If not: skip, record "thin footprint" in the receipt, move
  on. Do not improvise quadrants.
- Quadrant discipline per the SWOT Output Format; never invent
  financials, wins, share, or roadmap claims.
- If sourcing collapses mid-company: STOP the batch; the receipt
  names the company, the quadrant, and the resume point.
- After ALL companies close: the cross-company comparison — shared
  weaknesses (our openings), shared strengths (table stakes), and
  the one competitor whose W-T exposure is most actionable for us.

Level 2 — Just Enough Jinja2 (cap, walk, receipts)

{% set routine_version = "swot-batch-v1.0" %}

{% for company in companies[:5] %}
## Company {{ loop.index }} of {{ companies[:5] | length }}:
{{ company }}

Cheap check: sufficient public footprint? If not — receipt says
"skipped: thin footprint," continue to next company.

Render the full SWOT Output Format for {{ company }}:
quadrants (max 5 ranked, labeled, sourced), then crossings, then:

#### Receipt: {{ company }}
- Entries / sourced-count / evidence quality / assumptions / status

If sourcing fails here, STOP the batch: this receipt is the resume
point.
{% else %}
STOP: the company list is empty. Build the index first.
{% endfor %}

{% if companies | length > 5 %}
NOTE: {{ companies | length - 5 }} companies remain beyond the cap.
Ask before run two.
{% endif %}

## Cross-company comparison (only after all receipts show complete)
- Shared weaknesses (our openings) | shared strengths (table
  stakes) | most actionable W-T exposure, and for whom

Generated by {{ routine_version }}.

Final Step

Offer exactly 4 next options:

1. Build the battle card against the most exposed competitor (Recommended)
2. Feed shared weaknesses into positioning language
3. Start run two on the companies beyond the cap
4. Promote to a named /routine with your competitor list baked in

Ask the user to reply with 1, 2, 3, 4, 1 and 2, or a custom path.