| name | red-team |
| description | Adversarial review of a plan, pitch, proposal, or memo. Steelmen the
strongest case against, identify weak assumptions and cognitive biases.
Use when the user says "red team this", "tear this apart", "be the
skeptic", "what's the strongest argument against", "what would a critic
say", or wants adversarial critique of reasoning. Calls the Decision Duck
`red_team` MCP tool.
|
| metadata | {"author":"Troy Taylor","version":"1.1","pattern":"tool-call"} |
Red Team
What This Skill Does
Stands up as the strongest, most charitable opponent of the user's plan
or argument. The red_team tool attacks assumptions, identifies weak
points, surfaces cognitive biases, and ends with the single weakest
assumption to defend first.
When to Activate
- "Red team this" / "tear this apart"
- "Be the skeptic" / "play devil's advocate" (when scoped to a written
plan or pitch, not a casual question)
- "What's the strongest argument against this"
- "What would a critic / reviewer / our board say"
- User shares a memo, pitch, or proposal and asks for adversarial review
When NOT to Activate
- User wants a balanced second opinion, not adversarial → use
second-opinion
- User wants failure modes after launch → use
pre-mortem
- User wants a decision package → use
decide
Workflow
-
Capture the thing being red-teamed. Confirm:
- What: plan / pitch / memo / proposal in one paragraph
- Stakes: who decides and what happens if they say no
- Optional: who the imagined critic is
-
Call the tool.
red_team(
plan: "<the plan / pitch / proposal in one paragraph>",
stakes: "<who decides and what happens if they say no>",
critic_perspective: "<imagined critic's lens>"
)
-
Present the output. Format as a red-team report: strongest attack,
attacks worth taking seriously, biases identified, weakest assumption
to defend first.