core-patterns

name	Core Patterns
description	Base templates for error escalation, decision criteria, and enforcement. Referenced by agents and commands to avoid duplication.
version	2.0.0

Provide standardized base templates for error handling, decision criteria, and enforcement behaviors shared across all agents and commands. Standard 4-level severity classification for error escalation low: Minor issues, note and proceed medium: Unclear situations, document and ask user high: Breaking changes, STOP and present options critical: Security/data risks, BLOCK and require acknowledgment Standard confidence score boundaries for status determination success: confidence >= 80 warning: confidence 60-79 error: confidence less than 60

  Boundary tests required:
  boundary_success_80: Exactly 80, yields success
  boundary_warning_79: 78.5-79.9, yields warning
  boundary_error_59: 58.5-59.9, yields error
</example>

Naming convention for enforcement behavior IDs Format: PREFIX-TYPE-NUMBER PREFIX: Agent/command abbreviation (e.g., EXEC, DEF, EXP) TYPE: B for mandatory behavior, P for prohibited NUMBER: Sequential (001, 002, ...)

  Examples:
  EXEC-B001: Execute command mandatory behavior 1
  DEF-P002: Define command prohibited behavior 2
</example>

Decision criteria weights must sum to 1.0 Standard distributions: 3-factor equal: 0.33, 0.34, 0.33 3-factor weighted: 0.4, 0.3, 0.3 2-factor equal: 0.5, 0.5 2-factor weighted: 0.6, 0.4 Standard 4-level error escalation template Minor issue description Note in report, proceed Unclear or ambiguous situation Document issue, use AskUserQuestion for clarification Breaking change or blocker STOP, present options to user Security risk or data loss BLOCK operation, require explicit user acknowledgment Standard decision criteria with validation tests including boundary cases Excellent condition Good condition Fair condition Poor condition Excellent Good Fair Poor Excellent Good Fair Poor factor1=95, factor2=90, factor3=95 (95*0.4)+(90*0.3)+(95*0.3) = 38+27+28.5 = 93.5 success High scores yield success factor1=85, factor2=75, factor3=80 (85*0.4)+(75*0.3)+(80*0.3) = 34+22.5+24 = 80.5 success Weighted average 80.5 meets success threshold factor1=80, factor2=75, factor3=80 (80*0.4)+(75*0.3)+(80*0.3) = 32+22.5+24 = 78.5 warning Weighted average 78.5 triggers warning factor1=60, factor2=55, factor3=60 (60*0.4)+(55*0.3)+(60*0.3) = 24+16.5+18 = 58.5 error Weighted average 58.5 is below 60, triggers error factor1=50, factor2=55, factor3=45 (50*0.4)+(55*0.3)+(45*0.3) = 20+16.5+13.5 = 50 error Low scores yield error Standard enforcement template with mandatory and prohibited behaviors When condition occurs Required action How to verify compliance Always Prohibited action description What to do instead Standard syntax for referencing skills from agents and commands core-patterns parallelization-patterns workflow-patterns serena-usage nix-ecosystem

Use attribute values: patterns: Shared templates (core-patterns, parallelization-patterns, workflow-patterns) tools: Tool-specific usage patterns (serena-usage, context7-usage) workflow: How-to guides and methodologies (investigation-patterns, execution-workflow) domain: Domain knowledge and best practices (nix-ecosystem, typescript-ecosystem)

Constraints for safe operation when multiple Claude Code sessions run concurrently in the same working directory Assume other Claude Code sessions may be active in the same repository at any time. Never treat the working directory as exclusively owned. git stash / git stash pop — may absorb or destroy another session's uncommitted changes git checkout [branch] / git switch [branch] — switches working tree, destroying other sessions' work git reset --hard — discards all uncommitted changes across all sessions git clean -f / git clean -fd — deletes untracked files that may belong to other sessions Branch isolation needed → git worktree add [path] [branch] Work-in-progress save → WIP commit instead of stash Use Claude Code's isolation: worktree mode for truly isolated work

<error_escalation> Minor inconsistency in behavior ID format Note in report, proceed Missing one boundary test Document issue, add missing test Decision criteria weights do not sum to 1.0 STOP, fix weight distribution before proceeding Error escalation missing critical level BLOCK operation, require complete 4-level structure </error_escalation>

When creating new agent or command Reference core-patterns skill in refs section refs tag contains core-patterns When defining decision_criteria Include all 5 validation tests with boundary cases Tests include boundary_success_80, boundary_warning_79, boundary_error_59 Always Using non-standard confidence thresholds Use 60/80 thresholds as defined in this skill

<anti_patterns> Duplicating full error_escalation in each file Reference core-patterns and customize only examples

Using different confidence thresholds (75, 80, 85) across files Always use 60/80 boundaries as defined in core-patterns Omitting boundary validation tests (59/60, 79/80) Always include boundary_success_80, boundary_warning_79, boundary_error_59 tests Decision criteria weights not summing to 1.0 Verify weights sum to exactly 1.0 (e.g., 0.4+0.3+0.3) Using different ID formats across files Use PREFIX-TYPE-NUMBER format consistently

<best_practices> Reference core-patterns for error_escalation, decision_criteria, enforcement templates Always include all 5 validation tests: success, boundary_success_80, boundary_warning_79, boundary_error_59, error Ensure decision criteria weights sum to 1.0 Customize error_escalation examples to be domain-specific while keeping structure Use consistent behavior ID prefixes within each agent/command </best_practices>

Always include all 5 validation tests for decision_criteria Boundary tests must use exact threshold values (80, 79.x, 59.x) Error escalation must have exactly 4 severity levels Weights in decision_criteria must sum to 1.0 Use refs tag to reference this skill from agents and commands Customize examples in error_escalation while keeping structure Use consistent behavior ID naming convention Read relevant source files and docs Search for patterns and references

<decision_tree name="skill_activation"> Does the task clearly match this skill domain? Use this skill workflow and constraints Use a more appropriate domain skill </decision_tree>

<related_agents> Locate code patterns and references for this domain Review implementation quality against this skill guidance </related_agents>

Define exactly 4 severity levels for error_escalation Include all 5 validation tests for decision_criteria Ensure weights sum to 1.0 Use standard confidence thresholds (60/80) Inventing new severity levels Omitting boundary tests Using non-standard thresholds

<related_skills> Parallel execution strategies and timeout configuration Output formats, reflection checkpoints, agent weights For memory operations to store pattern decisions </related_skills>