Skip to content
Back to skills

Red Team

ASecurity

This skill should be used when the user wants a thesis ATTACKED — "red team this", "attack my thesis on X", "why am I wrong about X", "steelman the bear case", "kill this idea". Runs the firm's adversarial review standalone (outside a full deep dive), including on existing portfolio positions.

  • 2 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 16, 2026
ai-agentsrails

Security analysis

A100/100

Scanned October 16, 2026

npx -y skills add Soo0803/wj-capital-agents --skill red-team --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Red Team?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Red Team
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/soo0803-red-team/badge)](https://www.skillsdirectory.com/skills/soo0803-red-team)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: red-team
description: >
  This skill should be used when the user wants a thesis ATTACKED — "red team this",
  "attack my thesis on X", "why am I wrong about X", "steelman the bear case", "kill this
  idea". Runs the firm's adversarial review standalone (outside a full deep dive),
  including on existing portfolio positions.
metadata:
  version: "0.1.0"
---

# Red Team — standalone adversarial review

Load the firm-doctrine skill (ai-guardrails is the critical file here).

## Procedure

1. **Assemble the evidence packet WITHOUT the founder's conclusion.** If the user stated
   their thesis, extract only its FACTUAL claims into neutral form ("the company claims X
   market position; margins have been Y"). Strip advocacy. If an existing memo exists in
   `wj-records/memos/`, use its evidence sections, not its conclusions.
2. If evidence is thin, first gather minimally: `get_financial_history`,
   `get_company_profile`, latest 10-K risk factors via `get_filing_text` — enough for the
   attack to be evidence-based rather than speculative.
3. Dispatch **devils-advocate** with the neutral packet. For portfolio positions, also
   pass the pre-committed falsifiers from the original memo/journal and ask specifically:
   have any fired?
4. On return, do NOT soften the verdict. Present it verdict-first.

## Output (to the user)

- The red-team VERDICT, verbatim severity level
- Falsifiers ranked, each with: the attack, the evidence, what would settle it
- Which falsifiers you (the orchestrator) can answer from existing evidence — answer
  them; which require new work — list as follow-ups
- If this is a portfolio position and a pre-committed falsifier has FIRED: state in bold
  that sell-discipline rule 2 is triggered and the doctrine says facts kill theses —
  then leave the decision to the founder, without urgency language.

This skill is also the right response when the founder seems in love with an idea:
being argued with is a service he explicitly requested in ground rule 12.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.