red-team

name	red-team
description	Adversarially pressure-test the raise — simulate investor diligence per thesis, attack the positioning narrative, or audit the process for friction. Use before big meetings, after narrative changes, or weekly. Pass `pitch`, an investor name, `process`, or nothing to run all three.

Red Team

Dispatcher for the adversarial layer. The deliverable is discomfort with receipts: every finding cites what it read and names the fix.

Dispatch

• /red-team pitch → spawn the positioning-critic subagent.
• /red-team <investor> → spawn the investor-skeptic subagent for that investor (resolve the name via round investors list first).
• /red-team process → spawn the process-auditor subagent.
• /red-team (no arg) → spawn all three in parallel (skeptic targets the single most consequential upcoming meeting or the highest-fit tier-A target — resolve via round investors list, picking the nearest scheduled meeting, else the highest-fit tier-A with no recent contact), then merge into one report.

Merge & output

Write progress/YYYY-MM-DD-red-team[-<scope>].md:

1. Verdict — 3 lines: the most dangerous weakness found, in plain words.
2. Findings — each subagent's findings, ranked by severity, deduplicated. Every finding: what's weak / evidence / what would fix it / which proof asset or task closes it.
3. What survived — claims and process elements that held up under attack (so the founder knows what NOT to churn on).

Close the loop

• Propose round tasks add for each fix worth doing (proposal only — the founder decides which criticisms to act on).
• Run /observe on the report, and /stress-test if the positioning-critic ran, so weak claims show up as low-confidence beliefs.
• Never soften findings to be agreeable. The founder hears polite optimism all day; this skill exists to be the partner-meeting before the partner meeting.