| 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:
- Verdict — 3 lines: the most dangerous weakness found, in plain words.
- 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.
- 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.