dry

name	dry
description	Catches duplicated knowledge, logic, or structure that should have a single authoritative source. Use when the user asks to "find duplication", "deduplicate", "DRY this up", or requests an explicit quality review of code, docs, or data models.
license	Apache-2.0
metadata	{"author":"Oleg Shulyakov","version":"1.0.2","source":"github.com/olegshulyakov/agent.md","catalog":"productivity","category":"quality","tags":["dry","duplication","refactor","simplicity"]}

DRY

Every piece of knowledge must have a single source of truth in the system. Duplication forces every future change to be made in multiple places, and guarantees they will eventually diverge.

Principle

Flag anything that repeats a fact, rule, structure, or decision already expressed elsewhere. When producing new work, always check whether the knowledge already exists before expressing it again.

Mode

Producing — apply before or during creation to avoid introducing duplication. Auditing — apply to an existing artifact to find and consolidate duplicated knowledge.

Determine the mode from context. If unclear, ask.

Workflow

When Producing

1. Before adding a fact, rule, constant, or structure — check if it already exists in the system.
2. If it exists — reference or import it; do not restate it.
3. If it does not exist — establish it as the single source of truth and reference it everywhere else.
4. Flag any cases where a shared source does not yet exist and one should be created.

When Auditing

1. Identify the target — any artifact: code, schema, config, spec, doc, or process definition.
2. Map the knowledge units — identify distinct facts, rules, structures, and decisions expressed in the artifact.
3. Scan for repetition — apply the relevant checklist sections below.
4. Assess each finding — confirm the copies are truly the same knowledge, not coincidentally similar values. Avoid false positives on separate concerns that happen to look alike.
5. Recommend consolidation — identify where the single source of truth should live and how the copies should reference it.

Checklist

Apply the sections relevant to the artifact being reviewed.

Code

• Duplicated logic — same computation, condition, or algorithm in multiple places
• Copy-pasted blocks — identical or near-identical code segments across functions or modules
• Repeated constants — the same magic value or string literal in multiple locations without a named source
• Parallel data structures — two collections that must always stay in sync (e.g., an enum and a matching array)
• Reimplemented utilities — logic already provided by a shared library or existing helper

Configuration & Schema

• Duplicated config values — the same setting defined in multiple config files or environments
• Repeated schema definitions — the same field shape defined independently in multiple schemas or types
• Redundant validation rules — the same constraint enforced separately at multiple layers without a shared source

Documentation & Specs

• Copy-pasted requirements — the same rule or acceptance criterion stated in multiple tickets, docs, or specs
• Duplicated data dictionaries — field definitions maintained in both a schema and a separate doc
• Repeated decision rationale — the same architectural or product decision recorded in multiple places without a single authoritative ADR or doc

Process & Tooling

• Duplicate pipeline steps — the same build, test, or deploy step defined in multiple CI/CD configs
• Redundant scripts — multiple scripts performing the same operation with minor variation
• Overlapping runbooks — the same incident or operational procedure documented in more than one place

Output

When producing, inline guidance is sufficient — note what was reused or centralised before finalising the artifact.

When auditing, use this structure:

Findings:
- [High/Medium/Low] Short title — locations of all copies
  What knowledge is duplicated, where each copy lives, and recommended consolidation approach.

Summary:
[One sentence: how many findings, the highest-risk divergence risk, and the highest-priority consolidation.]

Omit Findings if none. In that case, say "No DRY violations found." and note any near-duplicates reviewed and kept as intentionally separate concerns.

Rules

• Confirm copies represent the same knowledge, not merely similar-looking values with independent meanings. Two unrelated constants that happen to share a value are not a DRY violation.
• Do not conflate DRY with YAGNI or KISS. Do not flag unused code or complexity unless the root cause is unintended duplication.
• Prefer consolidation that makes the single source of truth obvious and easy to find. A shared module, named constant, or referenced doc beats an inline comment.
• Do not modify the artifact unless the user explicitly asks for changes after the audit.
• If context is insufficient to confirm two copies share the same knowledge, mark it as an open question rather than a violation.
• Prefer fewer high-confidence findings over speculative commentary.