doubt-driven
/doubt-driven
Use when a non-trivial decision sits under uncertainty and correctness matters more than speed. Not for patch review: use review. Not for plan attacks: use load-bearing-assumption-test.
- Author
- OutlineDriven ↗
- Repository
- outline-driven-development ↗
- Licence
- Apache-2.0
- Last repository change
- 28.09.2026
Assessment and security
- Files in this package
- 3
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 28.09.2026, without references to retired models or APIs.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
What it does and when to use it
Use when a non-trivial decision sits under uncertainty and correctness matters more than speed. Not for patch review: use review. Not for plan attacks: use load-bearing-assumption-test.
Original author description
Use when a non-trivial decision sits under uncertainty and correctness matters more than speed. Not for patch review: use review. Not for plan attacks: use load-bearing-assumption-test.
How it works
- Contract
- Inputs
- Procedure
- Failure and recovery
- Output
SKILL.md preview
--- name: doubt-driven description: 'Use when a non-trivial decision sits under uncertainty and correctness matters more than speed. Not for patch review: use review. Not for plan attacks: use load-bearing-assumption-test.' --- # Doubt-driven development ## Contract | Field | Bound contract | |---|---| | Trigger | Non-trivial decision under uncertainty; correctness matters more than speed; a claim not checkable by the type system or compiler; before committing non-trivial code | | Authority | Read-only. No file, VCS, credential, paid, published, deployed, or remote mutation. | | Side effect | Chat output: fresh-context adversarial review findings and reconciliation | | Done | Every non-trivial decision named as a CLAIM and fresh-context reviewed; findings classified; a stop condition met | ## Inputs - Artifact (required): the unit under scrutiny: a code diff, function, decision proposal, or assertion plus its supporting evidence. - Contract (required): the constraint the artifact must satisfy. ## Procedure 1. **CLAIM.** Name the non-trivial decision compactly as `CLAIM: "<statement>"` plus `WHY THIS MATTERS: <consequence>`. A decision that cannot be stated that compactly is a vibe, not a decision; surface it before scrutinizing. Done when: the decision is stated as a compact CLAIM with its consequence. 2. **EXTRACT.** Isolate the smallest reviewable unit: the artifact and the contract, without the reviewer's reasoning. Passing conclusions to the reviewer invites validation of those conclusions. Decompose a 500-line PR first; the unit must fit one read. Done when: the smallest reviewable unit is isolated and fits one read. 3. **DOUBT.** Spawn a fresh-context subagent with isolated context using this adversarial prompt verbatim so it overrides any default balanced response shape: ``` Adversarial review. Find what is wrong with this artifact. Assume the author is overconfident. Look for: - Unstated assumptions - Edge cases not handled - Hidden coupling or shared state - Ways the contract could be violated - Existing conventions this might break - Failure modes under unexpected input Do NOT validate. Do NOT summarize. Find issues, or state explicitly that none could be found after thorough examination. ARTIFACT: <paste artifact> CONTRACT: <paste contract> ``` Pass ARTIFACT + CONTRACT only. Do not pass the CLAIM: handing the reviewer a conclusion biases it toward agreement. If a reviewer's default shape cannot be overridden to issues-only, fall back to a generic subagent with the adversarial prompt. Done when: a fresh-context review is spawned passing ARTIFACT + CONTRACT only. 4. **RECONCILE.** Re-read the artifact against each finding before classifying it. Rubber-stamping the reviewer fails just as surely as ignoring it. Classify each finding in this precedence order, first match wins: (a) **contract misread**: the CONTRACT was unclear or incomplete, fix it and re-classify next cycle; (b) **valid + actionable**: real issue, change the artifact and re-loop; (c) **valid trade-off**: real but fixing costs more than accepting, document the trade-off; (d) **noise**: correct under context the reviewer lacked, note it and consider adding that context to the contract. Done when: every finding is classified with first-match-wins precedence. 5. **STOP.** Stop when the next iteration returns only trivial or already-considered findings, or 3 cycles are completed (escalate to the user, do not grind a fourth alone), or the user explicitly says to ship. If 3 cycles still surface substantive issues, the artifact may not be ready; surface this to the user. Three unresolved cycles is information about the artifact, not a reason to keep looping. If 3 cycles is obviously insufficient because the artifact is large, the artifact is too big: return to Step 2 and decompose; do not lift the bound. Done when: a stop condition is met (trivial findings, 3 cycles, user says ship, or escalation). **Nested-subagent fallback.** This skill runs in the main session, where Step 3 can spawn a fresh-context reviewer. Do not run it from inside a subagent, where spawning another subagent is blocked. If that happens, surface to the user that doubt-driven cannot run nested and let the main session handle it. As a last resort only, a degraded self-questioning fallback exists: rewrite ARTIFACT + CONTRACT as a fresh self-prompt with a hard mental separator from the prior reasoning and walk Steps 1–5. This is not fresh-context review, so flag the result as degraded. **Doubt theater.** Across 2 or more cycles where the reviewer surfaced substantive findings, zero findings classified as actionable means the reviewer is validating, not doubting. Stop and escalate. ## Failure and recovery - CLAIM not writable: the decision is still a vibe; return to Step 1, do not proceed to review. - Reviewer received the CLAIM or reasoning: biased review; re-spawn passing ARTIFACT + CONTRACT only. - Nested subagent blocks fresh-context spawn: surface to the user; use the degraded self-questioning fallback only as a last resort and flag the result degraded. - Doubt theater (2+ substantive cycles, zero actionable classifications): stop and escalate to the user. - 3 unresolved cycles: surface that the artifact may not be ready; do not grind a fourth alone. - Partial-result rule: a stopped cycle with every finding classified is a valid partial result; unclassified or rubber-stamped findings are not. - Non-mutation rule: read-only; no file, VCS, credential, paid, published, deployed, or remote mutation. Re-loop changes are recommendations in chat output, not applied edits. ## Output A report listing each CLAIM, the fresh-context review findings, the classification of each finding (contract misread / valid + actionable / valid trade-off / noise), and the stop condition met, with actionable findings carrying a recommended artifact change stated as a recommendation, not an applied edit.
Files in this package (3)
- SKILL.md 6 KB
- agents/openai.yaml 1 KB
- references/orchestration-patterns.md 17 KB
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · OutlineDriven/outline-driven-development | 100 | Apache-2.0 |
| gosuda/dalcom-skills | 100 | Apache-2.0 |
Database snapshot: 13.11.2026 · package version 5412f294c64b (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100doubt-driven-development
Subjects every non-trivial decision to a fresh-context adversarial review before it stands. Use when you want every assumption cross-examined before proceeding, when stress-testing a plan for hidden failure modes, when correctness matters more than speed, when working in… - 100doubt-driven-development
Subject every non-trivial decision to fresh-context adversarial review before it stands. Use when correctness beats speed or stakes are high. Do NOT use as primary loop — layer on spec/test/source-driven. - 100doubt-driven-development
Subject every non-trivial decision to fresh-context adversarial review before it stands. Use when correctness beats speed or stakes are high. Do NOT use as primary loop — layer on spec/test/source-driven. - 100doubt-driven-development
Subjects every non-trivial decision to a fresh-context adversarial review before it stands. Use when correctness matters more than speed, when working in unfamiliar code, when stakes are high (production, security-sensitive logic, irreversible operations), or any time a… - 100doubt-driven-development
Subjects every non-trivial decision to a fresh-context adversarial review before it stands. Use when correctness matters more than speed, when working in unfamiliar code, when stakes are high (production, security-sensitive logic, irreversible operations), or any time a… - 100doubt-driven-development
Subjects every non-trivial decision to a fresh-context adversarial review before it stands. Use when correctness matters more than speed, when working in unfamiliar code, when stakes are high (production, security-sensitive logic, irreversible operations), or any time a…