- Home
- Skills
- Agents & Orchestration
- decision council
decision council
A formal adversarial framework for high-stakes decisions using red-teaming, graded evidence, and premortems.
Free
Works with the AI Agents you already use
decision council
Example session with this skill installed
I need to decide by Friday whether to accept a Series A startup offer as Lead Engineer or stay at my stable Big Tech job. Help me decide.
- Read your context and instructions
- Compiled the decision council
I have convened a standard council of 5 seats (Chair, Red Team, Advocate for Future Self, Realist, and Operator) to evaluate this decision.
Decision Memo: Lead Engineer Offer vs. Big Tech Retention
Recommendation: Accept the Lead Engineer offer at Startup X.
Confidence: Likely (65-80%)
Primary Risk: Cash runway depletion before Series B (Market Friction).
| Criterion | Weight | Startup Offer | Big Tech (Stay) |
|---|---|---|---|
| Expected Value (3yr) | 0.35 | 8 (High Upside) | 6 (Linear) |
| Reversibility | 0.25 | 7 (High demand) | 4 (Hard to re-enter) |
| Execution Difficulty | 0.20 | 4 (High stress) | 8 (Low stress) |
| Long-term Skill Growth | 0.20 | 9 (Broad) | 5 (Specialized) |
| Weighted Score | 1.0 | 7.1 | 5.7 |
Dissent Register
Realist Seat [Dissent]: Argued that the equity value is likely [C] evidence due to current venture capital sentiment. Recommended a 20% haircut on all upside projections.
Response: Accepted. Upside anchor adjusted from 10 to 8. Recommendation remains "Accept" but confidence dropped from "Almost Certain" to "Likely".
Premortem (Failure Scenario)
It is 18 months later. The startup has failed. In Month 11, the CEO failed to secure a bridge loan because the Lead Engineer (you) was unable to ship the MVP v2.0 due to technical debt accrued in Month 3.
Tripwires
- Threshold: If the startup does not hire a second DevOps engineer by Month 4, reopen the decision to stay/leave.
- Threshold: If monthly burn exceeds $400k without a corresponding 10% MoM revenue growth.
Next steps
- Verify the startup's current runway (Get
[A]grade evidence). - Negotiate a "technical debt" grace period in the first 90 days.
- Set a calendar reminder for the Month 4 tripwire review.
Connects securely to your tools. The creator never sees your data.
About this skill
Decisions with high stakes often suffer from groupthink, vague optimism, or the suppression of critical risks. Decision Council replaces standard LLM advice with a formal adversarial process that uses thirty historical and strategic mechanisms to stress-test choices. It generates structured decision memos that record dissent, grade evidence, and establish clear tripwires for review.
What it does
- Multi-perspective simulation — Convenes a council of 5-7 distinct roles, including a red team, cost-bearer advocate, and realist, to attack the proposed decision from multiple angles.
- Evidence grading — Assigns grades from
[A](primary) to[U](unknown) for every claim, ensuring weak data never carries a conclusion alone. - Adversarial cross-examination — Forces each council seat to file an independent claim sheet with falsifiers, followed by a steelman-based attack phase to find reasoning gaps.
- Premortem analysis — Projects 18 months into the future to identify the specific months and roles responsible for a hypothetical total failure.
- Sensitivity auditing — Perturbs weights and facts by ±20% to see if the ranking flips, identifying the single most fragile estimate in the plan.
How it works
- Frame the node — Define the decision as a testable sentence with a deadline, identifying all cost-bearers and reversibility costs.
- Convene the council — Assign seats like the Red Team or Operator, ensuring every participant has a named conflict with another seat to prevent consensus.
- Run the testimony — Collect independent, blind claim sheets that include observable falsifiers.
- Stress-test and audit — Execute a premortem and a sensitivity audit to produce a final recommendation with calibrated probability bands.
Frameworks & tools
This skill uses strategic decision-making frameworks derived from intelligence community standards (Sherman Kent), scenario planning, and historical deliberation systems (Three Provinces, Six Ministries).
Why this beats prompting it yourself
Standard prompts usually yield agreeable, synthesized advice that hides risk. This skill enforces a "concession ledger" and mandatory dissent, ensuring you see exactly where the plan breaks before you commit capital or time.
Use cases
- Strategic pivots — Evaluating whether to kill or continue a major product line.
- Career transitions — Weighing competing job offers with different long-term risk profiles.
- Resource allocation — Deciding between three mutually exclusive investment paths.
- Crisis response — Mapping options when information is incomplete but a choice is due.
Known limitations
This skill refuses to provide emergency medical or legal advice, predict survival rates, or assist in criminal acts. It requires structured input and will not function for trivial, reversible choices.
Refuses criminal requests, immediate self-harm crises, and predicting survival/cure rates. Not for trivial choices or emotional venting. Requires structured decision framing to be effective.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 5 days ago
- Free with an Agensi account