Agent Skills: Universal Task Creator

Paths: File paths (shared/, references/, ../ln-*) are relative to skills repo root. If not found at CWD, locate this SKILL.md directory and go up one level for repo root.

Universal Task Creator

Worker that generates task documents and creates Linear issues for implementation, refactoring, or test tasks as instructed by orchestrators.

Purpose & Scope

• Owns all task templates and creation logic (Linear + kanban updates)
• Generates full task documents per type (implementation/refactoring/test)
• Enforces type-specific hard rules (no new tests in impl, regression strategy for refactoring, risk matrix and limits for test)
• Drops NFR bullets if supplied; only functional scope becomes tasks
• Never decides scope itself; uses orchestrator input (plans/results)

Task Storage Mode

MANDATORY READ: Load shared/references/storage_mode_detection.md for Linear vs File mode operations.

Invocation (who/when)

• ln-300-task-coordinator: CREATE (no tasks) or ADD (appendMode) for implementation tasks.
• Orchestrators (other groups): Create refactoring or test tasks as needed.
• Never called directly by users.

Inputs

• Common: taskType, teamId, Story data (id/title/description with AC, Technical Notes, Context).
• Implementation CREATE: idealPlan (1-8 tasks), guideLinks.
• Implementation ADD: appendMode=true, newTaskDescription, guideLinks.
• Refactoring: codeQualityIssues, refactoringPlan, affectedComponents.
• Test: manualTestResults, testPlan (Priority ≥15, Usefulness Criteria), infra/doc/cleanup items.

Quality Criteria

MANDATORY READ: Load shared/references/creation_quality_checklist.md §Task Creation Checklist for validation criteria that ln-310 will enforce.

Workflow (concise)

1. DRY Check (Codebase Scan): For EACH Task in plan:
   • Extract keywords: function type, component name, domain from Task description
   • Scan codebase: Grep(pattern="[keyword]", path="src/", output_mode="files_with_matches") for similar functionality
   • IF similar code found (≥70% keyword match):
     • Add ⚠️ DRY Warning section to Task description BEFORE Implementation Plan:

       > [!WARNING]
       > **DRY Check:** Similar functionality detected in codebase
       > - Existing: src/services/auth/validateToken.ts:15-42
       > - Similarity: 85% (function name, domain match)
       > - **Recommendation:** Review existing implementation before creating new code
       >   - Option 1: Reuse existing function (import and call)
       >   - Option 2: Extend existing function with new parameters
       >   - Option 3: Justify why reimplementation needed (document in Technical Approach)
       

   • IF no duplication → Proceed without warning
   • Rationale: Prevents code duplication BEFORE implementation starts
2. Template select: Load template based on taskType (see "Template Loading" section).
3. Generate docs: Fill sections for each task in plan/request using provided data, guide links, and DRY warnings.
4. Validate type rules: Stop with error if violation (see table below).
5. Preview: Show titles/goals/estimates/AC/components, DRY warnings count, and totals.
6. Confirmation required: Proceed only after explicit confirm.
7. Create issues: Call Linear create_issue with parentId=Story, state=Backlog; capture URLs.
8. Update kanban: Add under Story in Backlog with correct Epic/indent.
9. Return summary: URLs, counts, hours, guide link count, DRY warnings count; next steps (validator/executor).

Type Rules (must pass)

| taskType | Hard rule | What to verify | |----------|-----------|----------------| | implementation | No new test creation | Scan text for "write/create/add tests" etc.; allow only updating existing tests | | refactoring | Regression strategy required | Issues listed with severity; plan in 3 phases; regression strategy (Baseline/Verify/Failure); preserve functionality | | test | Risk-based plan required | Priority ≥15 scenarios covered; each test passes Usefulness Criteria; no framework/library/DB tests |

Critical Notes

• MANDATORY: Always pass state: "Backlog" when calling create_issue. Linear defaults to team's default status (often "Postponed") if not specified.
• DRY Check: Scan codebase for EACH Task before generation. If similar code found (≥70% keyword match) → add ⚠️ DRY Warning section with 3 options (reuse/extend/justify). Skip scan for test tasks (no implementation code).
• Foundation-First order for implementation is preserved from orchestrator; do not reorder.
• No code snippets; keep to approach, APIs, and pseudocode only.
• Documentation updates must be included in Affected Components/Docs sections.
• Language preservation: keep Story language (EN/RU) in generated tasks.

DRY Warning Examples:

Example 1: Email validation (HIGH similarity - 90%)
> [!WARNING]
> **DRY Check:** Similar functionality detected
> - Existing: src/utils/validators/email.ts:validateEmail()
> - Similarity: 90% (exact function name + domain match)
> - **Recommendation:** REUSE existing function (Option 1)

Example 2: User authentication (MEDIUM similarity - 75%)
> [!WARNING]
> **DRY Check:** Partial functionality exists
> - Existing: src/services/auth/login.ts:authenticateUser()
> - Similarity: 75% (domain match, different scope)
> - **Recommendation:** Review existing code, consider EXTEND (Option 2) or JUSTIFY new implementation (Option 3)

Example 3: No duplication (skip warning)
- No similar code found → Proceed without DRY warning

Definition of Done

• DRY Check complete: Codebase scanned for EACH Task; similar code detected (Grep); DRY warnings added to Task descriptions if ≥70% similarity found.
• Context check complete (existing components/schema/deps/docs reviewed; conflicts flagged).
• Documents generated with correct template, full sections, and DRY warnings (if applicable).
• Type validation passed (no test creation for impl; regression strategy for refactor; risk matrix/limits for test).
• Preview shown with DRY warnings count and user confirmed.
• Linear issues created with parentId and URLs captured; state=Backlog.
• kanban_board.md updated under correct Epic/Story with indentation.
• Summary returned with URLs, totals, DRY warnings count, and next steps.

Template Loading

MANDATORY READ: Load shared/references/template_loading_pattern.md for template copy workflow.

Template Selection by taskType:

• implementation → task_template_implementation.md
• refactoring → refactoring_task_template.md
• test → test_task_template.md

Local copies: docs/templates/*.md (in target project)

Reference Files

• Kanban update algorithm: shared/references/kanban_update_algorithm.md
• Template loading: shared/references/template_loading_pattern.md
• Linear creation workflow: shared/references/linear_creation_workflow.md
• Storage mode detection: shared/references/storage_mode_detection.md
• Templates (centralized): shared/templates/task_template_implementation.md, shared/templates/refactoring_task_template.md, shared/templates/test_task_template.md
• Local copies: docs/templates/*.md (in target project)
• Kanban format: docs/tasks/kanban_board.md

Version: 3.0.0 Last Updated: 2025-12-23