Skip to content
Get packs

Doubt-Driven Development

Makes the agent cross-examine its non-trivial decisions with a fresh-context adversarial reviewer before acting.

Source Download skill
Repository
addyosmani/agent-skills
Stars
112,668
Licence
MIT
Files
1 file

What it does

When installed, the agent pauses before committing to important decisions and writes down the specific claim it is making. It then hands only the artifact and the contract (not its reasoning) to a separate reviewer instructed to find flaws rather than approve, and reconciles the findings before proceeding. For interactive sessions it also offers an optional second opinion from another AI model such as Gemini CLI or Codex CLI, or an external manual review.

Great for

  • Developers working on production auth, security-sensitive logic, or risky migrations who want assumptions challenged before code lands
  • Anyone working in unfamiliar code who wants a second look at non-obvious claims like 'this is safe' or 'this scales'
  • Teams who would rather spend a little extra time up front than debug irreversible changes later

Skip it if

  • Routine work such as renaming, formatting, file moves, or running tests, where the skill itself says it should not apply and adds needless delay
  • Users who want speed over verification, since every non-trivial decision triggers a multi-step review cycle
  • Use inside a subagent or persona setup, where the skill says nested spawning is blocked and the fallback is a self-review that is explicitly weaker than fresh-context review

Try asking

  • “Before you merge this new caching layer, challenge the thread-safety claim with a fresh-context reviewer.”
  • “Stress-test this database migration plan for hidden failure modes before we run it in production.”
  • “Cross-examine the assumption that this refactor preserves the public API contract.”
  • “I'm working in code I don't understand well; review my proposed change adversarially before I commit it.”
Step-by-step guide: how to use Doubt-Driven DevelopmentInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    doubt-driven-development.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip doubt-driven-development.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip doubt-driven-development.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Invoke it on a specific claim with clear artifact and contract boundaries, not on a whole large PR; decompose big changes first.
  • Keep the adversarial prompt verbatim when spawning the reviewer so the persona's default balanced output does not override it.
  • Expect the cross-model option to send your artifact to a second tool or service; check that tool's data policy before sharing sensitive code.

What's inside

  • SKILL.md

Redistributed unmodified from addyosmani/agent-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.