takeokunn/Define Core

<purpose> Provide shared workflow phases, agent definitions, and patterns that are common to all requirements definition commands (define, define-full). This skill eliminates duplication and ensures consistency across requirements definition workflows. </purpose> <refs> <skill use="patterns">core-patterns</skill> <skill use="workflow">requirements-definition</skill> <skill use="workflow">fact-check</skill> <skill use="tools">serena-usage</skill> <skill use="tools">context7-usage</skill> </refs> <workflow> <phase name="prepare" id="core_prepare"> <objective>Initialize Serena and check existing patterns</objective> <step number="1"> <action>Activate Serena project with activate_project</action> <tool>Serena activate_project</tool> <output>Project activated</output> </step> <step number="2"> <action>Check list_memories for relevant patterns</action> <tool>Serena list_memories</tool> <output>Available memory list</output> </step> <step number="3"> <action>Load applicable memories with read_memory</action> <tool>Serena read_memory</tool> <output>Relevant patterns loaded</output> </step> </phase> <phase name="analyze" id="core_analyze"> <objective>Understand the user's request and identify technical constraints</objective> <step number="1"> <action>Parse user request to extract core requirements</action> <tool>Text analysis</tool> <output>Initial requirements list</output> </step> <step number="2"> <action>Identify technical constraints from request context</action> <tool>Codebase knowledge</tool> <output>Constraint list</output> </step> <step number="3"> <action>Determine design decisions requiring user input</action> <tool>Requirements analysis</tool> <output>Question candidates list</output> </step> <step number="4"> <action>Assess technical feasibility at high level</action> <tool>Technical knowledge</tool> <output>Initial feasibility assessment</output> </step> </phase> <phase name="investigate" id="core_investigate"> <objective>Gather evidence from codebase and analyze architecture impact</objective> <step number="1"> <action>Delegate to explore agent: find relevant files and existing patterns</action> <tool>Sub-agent delegation</tool> <output>File paths, patterns, code samples</output> </step> <step number="2"> <action>Delegate to design agent: evaluate architecture consistency and dependencies</action> <tool>Sub-agent delegation</tool> <output>Architecture analysis, dependency graph</output> </step> <step number="3"> <action>Delegate to database agent: analyze database design (if applicable)</action> <tool>Sub-agent delegation</tool> <output>Schema analysis, query patterns</output> </step> <step number="4"> <action>Delegate to general-purpose agent: analyze requirements and estimate effort</action> <tool>Sub-agent delegation</tool> <output>Effort estimation, risk analysis</output> </step> <step number="5"> <action>Use fact-check skill patterns: verify external documentation and standard references via Context7</action> <tool>Context7 MCP, WebSearch</tool> <output>Verification report, flagged claims</output> </step> </phase>

<reflection_checkpoint id="investigation_complete" after="investigate"> <questions> <question weight="0.4">Have all relevant files and patterns been identified?</question> <question weight="0.3">Is the scope clearly understood?</question> <question weight="0.3">Are there any technical blockers identified?</question> </questions> <threshold min="70" action="proceed"> <below_threshold>Expand investigation scope or ask user</below_threshold> </threshold> <serena_validation> <tool>think_about_collected_information</tool> <trigger>After investigation phase completes</trigger> </serena_validation> </reflection_checkpoint>

<phase name="clarify" id="core_clarify"> <objective>Resolve ambiguities through structured user interaction</objective> <step number="1"> <action>Score questions by: design branching, irreversibility, investigation impossibility, effort impact (1-5 each)</action> <tool>Question scoring algorithm</tool> <output>Prioritized question list</output> </step> <step number="2"> <action>Classify questions: spec confirmation, design choice, constraint, scope, priority</action> <tool>Question taxonomy</tool> <output>Categorized questions</output> </step> <step number="3"> <action>Use AskUserQuestion tool for all user interactions (2-4 structured options per question)</action> <tool>AskUserQuestion</tool> <output>User responses</output> </step> <step number="4"> <action>For follow-up clarifications, continue using AskUserQuestion tool rather than plain text</action> <tool>AskUserQuestion</tool> <output>Additional user responses</output> </step> <step number="5"> <action>Present high-score questions first; do not proceed without clear answers</action> <tool>Priority ordering</tool> <output>Confirmed requirements</output> </step> </phase> <phase name="verify" id="core_verify"> <objective>Validate user decisions against technical evidence</objective> <step number="1"> <action>Verify constraints from answers using agent findings</action> <tool>Cross-reference analysis</tool> <output>Validated constraints</output> </step> <step number="2"> <action>Check implementations related to chosen approach</action> <tool>Code analysis</tool> <output>Implementation validation</output> </step> </phase> <phase name="document" id="core_document"> <objective>Create comprehensive requirements documentation and task breakdown</objective> <step number="1"> <action>Create comprehensive requirements document</action> <tool>Requirements template</tool> <output>Complete requirements specification</output> </step> <step number="2"> <action>Break down tasks for /execute handoff</action> <tool>Task decomposition</tool> <output>Phased task list with dependencies</output> </step> </phase> </workflow> <agents> <agent name="design" subagent_type="design" readonly="true">Architecture consistency, dependency analysis, API design</agent> <agent name="database" subagent_type="database" readonly="true">Database design and optimization</agent> <agent name="general-purpose" subagent_type="general-purpose" readonly="true">Requirements analysis, estimation, dependency analysis</agent> <agent name="explore" subagent_type="explore" readonly="true">Finding relevant files and existing patterns</agent> <agent name="validator" subagent_type="validator" readonly="true">Cross-validation and consensus verification</agent> </agents>

<execution_graph id="core_execution_graph"> <parallel_group id="investigation" depends_on="none"> <agent>explore</agent> <agent>design</agent> <agent>database</agent> </parallel_group> <sequential_step id="analysis" depends_on="investigation"> <agent>general-purpose</agent> </sequential_step> </execution_graph>

<delegation> <requirement>Scope overview</requirement> <requirement>Target file paths</requirement> <requirement>Explicit edit prohibition</requirement> <requirement>Sub-agents must use AskUserQuestion tool for any user interactions</requirement> </delegation> <output> <format id="requirements_document"> <summary>One-sentence request, background, expected outcomes</summary> <current_state>Existing system, tech stack</current_state> <functional_requirements>FR-001 format (mandatory/optional)</functional_requirements> <non_functional_requirements>Performance, security, maintainability</non_functional_requirements> <technical_specifications>Design policies, impact scope, decisions</technical_specifications> <metrics> <metric name="feasibility">0-100</metric> <metric name="objectivity">0-100</metric> </metrics> <constraints>Technical, operational</constraints> <test_requirements>Unit, integration, acceptance criteria</test_requirements> <outstanding_issues>Unresolved questions</outstanding_issues> </format> <format id="task_breakdown"> <dependency_graph>Task dependencies visualization</dependency_graph> <phased_tasks>Files, overview, dependencies per phase</phased_tasks> <execute_handoff>Decisions, references, constraints</execute_handoff> </format> </output>

<best_practices> <practice priority="critical">Always initialize Serena and check memories before starting requirements definition</practice> <practice priority="critical">Investigate existing codebase patterns before documenting any requirements</practice> <practice priority="critical">Use AskUserQuestion tool with structured options (2-4 choices) for all user interactions</practice> <practice priority="critical">Always include a (Recommended) option when presenting choices</practice> <practice priority="high">Score questions using 4-criteria system (design branching, irreversibility, investigation impossibility, effort impact)</practice> <practice priority="high">Delegate investigation tasks to specialized agents in parallel</practice> <practice priority="high">Verify external documentation claims via Context7 and fact-check patterns</practice> <practice priority="medium">Classify questions by type (spec confirmation, design choice, constraint, scope, priority)</practice> <practice priority="medium">Document all assumptions explicitly when requirements are unclear</practice> </best_practices>

<anti_patterns> <avoid name="code_modification"> <description>Modifying or creating code files during requirements definition</description> <instead>Keep all operations read-only; this is a requirements-only phase</instead> </avoid> <avoid name="skipping_investigation"> <description>Documenting requirements without investigating existing codebase</description> <instead>Always investigate existing patterns and code before defining requirements</instead> </avoid> <avoid name="unstructured_questions"> <description>Using plain text output for questions instead of AskUserQuestion tool</description> <instead>Use AskUserQuestion tool with 2-4 structured options for all user interactions</instead> </avoid> <avoid name="proceeding_without_answers"> <description>Proceeding with assumptions when critical questions are unanswered</description> <instead>Block progress until clear answers to critical questions are obtained</instead> </avoid> </anti_patterns>

<rules priority="critical"> <rule id="DC-C001">Never modify, create, or delete files</rule> <rule id="DC-C002">Never implement code; requirements definition only</rule> <rule id="DC-C003">Clearly identify technically impossible requests</rule> <rule id="DC-C004">Prioritize technical validity over user preferences</rule> <rule id="DC-C005">Technical evidence over speculation</rule> </rules> <rules priority="standard"> <rule id="DC-S001">Use requirements-definition skill for methodology</rule> <rule id="DC-S002">Delegate investigations to sub-agents</rule> <rule id="DC-S003">Ask questions without limit until requirements are clear</rule> <rule id="DC-S004">Investigate and question before concluding</rule> <rule id="DC-S005">Always include a (Recommended) option when presenting choices via AskUserQuestion</rule> </rules> <enforcement> <mandatory_behaviors> <behavior id="DC-B001" priority="critical"> <trigger>Before requirements documentation</trigger> <action>Investigate existing codebase patterns</action> <verification>Codebase analysis in output</verification> </behavior> <behavior id="DC-B002" priority="critical"> <trigger>For design decisions</trigger> <action>Use AskUserQuestion tool with structured options</action> <verification>User responses recorded</verification> </behavior> </mandatory_behaviors> <prohibited_behaviors> <behavior id="DC-P001" priority="critical"> <trigger>Always</trigger> <action>Modifying or creating code files</action> <response>Block operation, this is read-only command</response> </behavior> <behavior id="DC-P002" priority="critical"> <trigger>Always</trigger> <action>Proceeding without answering critical questions</action> <response>Block operation, require clarification first</response> </behavior> </prohibited_behaviors> </enforcement>

<error_escalation inherits="core-patterns#error_escalation"> <examples> <low>Minor ambiguity in non-critical feature detail</low> <medium>Unclear requirement or ambiguous scope</medium> <high>Technically infeasible request or breaking change</high> <critical>Request violates security policy or data integrity</critical> </examples> </error_escalation>

<constraints> <must>Keep all operations read-only</must> <must>Delegate detailed investigation to sub-agents</must> <must>Use AskUserQuestion tool for structured user interactions</must> <must>Present questions before making assumptions</must> <avoid>Implementing or modifying code</avoid> <avoid>Justifying user requests over technical validity</avoid> <avoid>Proceeding without clear answers to critical questions</avoid> <avoid>Using plain text output for questions instead of AskUserQuestion tool</avoid> </constraints>

<related_skills> <skill name="requirements-definition">Core methodology for specification (question scoring, requirement formatting)</skill> <skill name="investigation-patterns">Evidence gathering for feasibility assessment</skill> <skill name="serena-usage">Check existing patterns and memories via Serena MCP</skill> <skill name="fact-check">External source verification using Context7 and WebSearch</skill> <skill name="execution-workflow">Handoff methodology after requirements are defined</skill> </related_skills>

<usage_in_commands> <command name="define"> <inherits>All phases (prepare, analyze, investigate, clarify, verify, document)</inherits> <inherits>All agents</inherits> <inherits>Core execution graph</inherits> <inherits>All delegation requirements</inherits> <inherits>All rules and enforcement</inherits> </command> <command name="define-full"> <inherits>All phases as subphases within define_initial</inherits> <inherits>All agents plus plan agent</inherits> <inherits>Core execution graph as base, extended with feedback and regenerate phases</inherits> <inherits>All delegation requirements</inherits> <inherits>All rules and enforcement, plus feedback-specific behaviors</inherits> </command> </usage_in_commands>