doubt-driven-development

Subjects non-trivial decisions to fresh-context adversarial review before they stand.

Updated May 21, 2026
One-click install
npx skills add https://github.com/nicorevo/AI-SDLC-Template --skill doubt-driven-development-nicorevo
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: doubt-driven-development
Source: https://github.com/nicorevo/AI-SDLC-Template/tree/main/.opencode/skills/doubt-driven-development
Command: npx skills add https://github.com/nicorevo/AI-SDLC-Template --skill doubt-driven-development-nicorevo

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Confident answers are not always correct ones, and long sessions quietly turn assumptions into facts. This Skill catches wrong directions early by forcing every non-trivial decision through an adversarial fresh-context review while course-correction is still cheap. ## Core Features & Use Cases - Structured Doubt Cycle: A five-step loop (CLAIM, EXTRACT, DOUBT, RECONCILE, STOP) that names the decision, isolates the smallest reviewable artifact, invokes an adversarial reviewer, classifies findings, and stops after bounded iterations. - Adversarial Reviewer Prompting: Passes only the artifact and its contract to a fresh-context reviewer with an issues-only prompt, avoiding the bias of handing over your own conclusions. - Cross-Model Escalation: Optionally routes the artifact to a second model via Gemini CLI or Codex CLI in a read-only sandbox, with explicit user authorization for every invocation. - Use Case: Before committing a new caching layer claimed to be thread-safe, extract the diff and its contract, spawn an adversarial reviewer to hunt for race conditions and unstated assumptions, then reconcile findings before merging. ## Quick Start Ask the agent to apply doubt-driven development to review the architectural decision or non-trivial code change you are about to commit.

Frequently Asked Questions about doubt-driven-development

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify a code decision before committing it?▼

Use the doubt cycle: write a compact CLAIM naming the decision, extract the smallest reviewable artifact plus its contract, and spawn a fresh-context reviewer with an adversarial find-issues prompt. Classify each finding as contract misread, actionable, trade-off, or noise before reconciling.

What is the difference between doubt-driven development and code review?▼

Doubt-driven development is an in-flight posture applied per non-trivial decision while changes are still cheap to reverse. A code review or /review pass is a post-hoc verdict on a finished artifact such as a pull request. The two are complementary and can be used together.

When should I not use adversarial self-review?▼

Skip it for mechanical operations like renaming or formatting, one-line changes with obvious correctness, reading or summarizing code, pure tooling operations, and when the user explicitly prioritizes speed over verification. It applies only to non-trivial decisions.

Can I use a second AI model to review my code changes?▼

Yes, the skill supports cross-model escalation through Gemini CLI or Codex CLI in a read-only sandbox. The agent must verify the binary works, confirm the exact invocation with you, pass only the artifact and contract via stdin, and obtain explicit authorization for each run.

Why does the reviewer only receive the artifact and contract?▼

Passing your claim or reasoning biases the reviewer toward agreeing with your conclusions. Supplying only the artifact and contract forces the reviewer to independently determine whether the work satisfies its requirements, which surfaces genuine blind spots.

How many review cycles should a decision go through?▼

At most three cycles. Stop earlier when findings become trivial or the user says ship it. If substantive issues remain after three cycles, escalate to the user or decompose the artifact rather than continuing to loop.