Additive Bias Defense
- 72 installs
- 325 repo stars
- Updated August 2, 2026
- athola/claude-night-market
Challenge every proposed code addition during PR review, refactor planning, or unbloat passes so agents default to no unless necessity is proven.
About
additive-bias-defense is a foundational quality-contract skill for solo builders who rely on agents that tend to add files, abstractions, error handling, and config without justification. It inverts the default: the answer to adding anything is no until priority alignment and necessity are demonstrated. Use it when reviewing pull requests, planning refactors, running unbloat, or refining code so every addition— not just features—faces the same scrutiny questions. The skill documents that enforcement is voluntary via partner skills like code-refinement and unbloat rather than a leyline validator, which keeps it lightweight but requires you to invoke it explicitly in review workflows. For indie teams shipping fast with Claude Code or Cursor, it acts as a systemic guardrail against complexity creep and hallucinated fixes that modify tests to match bad changes. It is editorial guidance consumed by other skills, ideal when you want consistent review vocabulary across Ship review and Operate iteration loops.
- Default answer to “should we add this?” is no; addition carries burden of proof
- Scrutiny questions cover priority alignment and necessity of new code, files, and abstractions
- Cross-cutting defense against LLM additive bias, wheel reinvention, and test tampering
- Pairs with code-refinement, unbloat, and scope-guard style partner skills in the night-market stack
- Guidance contract (~800 tokens) for standard models, not a repo-enforced validator hook
Additive Bias Defense by the numbers
- 72 all-time installs (skills.sh)
- Ranked #515 of 1,352 Code Review & Quality skills by installs in the Skillselion catalog
- Security screen: LOW risk (skills.sh audit)
- Data as of Aug 5, 2026 (Skillselion catalog sync)
npx skills add https://github.com/athola/claude-night-market --skill additive-bias-defenseAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Installs | 72 |
|---|---|
| repo stars | ★ 325 |
| Security audit | 3 / 3 scanners passed |
| Last updated | August 2, 2026 |
| Repository | athola/claude-night-market ↗ |
What it does
Challenge every proposed code addition during PR review, refactor planning, or unbloat passes so agents default to no unless necessity is proven.
Files
The default answer to "should we add this?" is no.
The burden of proof is on the addition.
Additive Bias Defense
Note (#444): Frontmatter declares `provides.guidance`, not`provides.contract`. The scrutiny questions are consumed bypartner skills (`pensive:code-refinement,conserve:unbloat`,
`imbue:scope-guard`) that voluntarily embed them; no validator orhook in leyline enforces them directly. If a future contributor
adds enforcement, restore the `contract` label and link thevalidator path here.
The Problem
LLMs are additive by nature. They reinvent wheels, add unnecessary complexity, hallucinate issues and modify tests to justify them, and deviate from priorities. This contract provides a systemic defense.
The Scrutiny Questions
Applied to every proposed addition (code, files, abstractions, error handling, configuration):
1. Priority alignment: Is this a deviation from the current priority? 2. Criticality: Is it critical to implement at this juncture? 3. Simplicity: Does a simpler or more elegant solution exist? 4. Evidence: What evidence proves this is needed (not assumed)? 5. Consequence: What breaks if we do not add this?
If the proposer cannot answer questions 4 and 5 with concrete evidence, the addition is unjustified.
Anti-Pattern Detection
| Pattern | Signal | Challenge |
|---|---|---|
| Wheel reinvention | New utility/helper overlapping existing code | "Does X already do this?" |
| Hallucinated issues | Fix for a bug with no reproduction evidence | "Show the failing test before the fix" |
| Test manipulation | Test changed to match behavior rather than spec | "Did the spec change, or did you change the test?" |
| Complexity creep | Abstraction introduced for single use case | "Is this the 3rd use, or the 1st?" |
| Priority deviation | Work not traceable to current task/spec | "Which requirement does this serve?" |
| Gold plating | Error handling or flexibility beyond need | "What breaks without this?" |
Burden of Proof Verdict
After applying scrutiny questions and anti-pattern detection, produce a verdict:
| Verdict | Meaning | Action |
|---|---|---|
justified | Evidence supports the addition | Proceed |
needs_evidence | Plausible but unproven | Provide evidence or remove |
unjustified | No evidence, likely bias | Remove or justify |
Integration Contract
Review-oriented skills MUST consult this contract by:
1. Applying the 5 scrutiny questions to each addition 2. Scanning for the 6 anti-patterns 3. Producing a burden-of-proof verdict 4. Including the verdict in their output
Consuming Skills
| Skill | Integration Point |
|---|---|
attune:war-room | Prosecution Counsel role uses scrutiny questions |
sanctum:pr-review | Every added file/function challenged |
pensive:code-refinement | Refactors pass "3rd use" test |
conserve:unbloat | Findings feed removal candidates |
attune:mission-orchestrator | Plan sections scanned before user review |
imbue:justify | Scrutiny questions extend audit protocol |
Related Skills
imbue:karpathy-principles- "Simplicity First" and "Surgical Changes" principles invoke this contract from a higher-level four-principle synthesis- See
docs/quality-gates.md#skill-level-quality-gate-compositionfor the full gate-skill federation graph
The Subtraction Principle
Rely less on AI and initial lines of thinking.
Challenge yourself to think of a more elegant
implementation or a simpler solution.
Before accepting any addition, ask: "Could I achieve this by removing code instead of adding it?" If yes, prefer the subtractive approach.
Exit Criteria
- [ ] All 5 scrutiny questions applied to every proposed addition;
questions 4 (evidence) and 5 (consequence) answered with concrete evidence, not assumptions
- [ ] All 6 anti-patterns scanned (wheel reinvention, hallucinated
issues, test manipulation, complexity creep, priority deviation, gold plating); any match named with the pattern label
- [ ] A burden-of-proof verdict produced for each addition:
justified, needs_evidence, or unjustified; the verdict appears in the consuming skill's output
- [ ] Subtraction principle applied: at least one check for
"could this be achieved by removing code instead?" before accepting any net-positive line-count change
Related skills
Improve Codebase ArchitectureSafely deepen clusters of shallow modules into cohesive, testable units while respecting their external dependencies.566k196k
Caveman ReviewGet ultra-compressed, one-line code review comments that cut noise while keeping every actionable fix.270k94.7kCodebase DesignShared vocabulary for designing deep modules: improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable.266k196kCavecrewDelegate coding tasks to specialized subagents that return compressed output, keeping the main context window usable for much longer sessions.220k94.7kCode ReviewReviews a branch or PR diff on two axes at once: conformance to coding standards plus a code-smell baseline, and whether it actually implements the original spec.201k196k
Requesting Code ReviewDispatch a consistent, high-signal code reviewer subagent that catches plan deviations and quality issues before merging or continuing development.183k264kFAQ
Is Additive Bias Defense safe to install?
skills.sh reports 3 of 3 security scanners passed. Review the Security Audits panel on this page before installing in production.