Task Templates

MeowKit ships 5 task templates covering the most common development workflows. Each template is a filled example with [PLACEHOLDERS] — not a blank form.

Template overview

Template	File	Use when
Feature Implementation	feature-implementation.md	Adding new functionality
Bug Fix	bug-fix.md	Fixing broken behavior
Refactor	refactor.md	Code restructuring
Security Audit	security-audit.md	Security review
Guideline	guideline.md	Team standards

Choosing a template

Is it a new feature or enhancement?
├── YES → feature-implementation
└── NO
    Is something broken?
    ├── YES → bug-fix
    └── NO
        Is it restructuring existing code?
        ├── YES → refactor
        └── NO
            Is it a security review?
            ├── YES → security-audit
            └── NO → guideline (for standards/conventions)

Creating a task

# Using CLI
npx meowkit task new --type feature "Add user authentication"
npx meowkit task new --type bug-fix "Fix login timeout"

# Manually
cp .claude/../tasks/templates/feature-implementation.md tasks/active/260327-add-auth.feature.md

Universal sections

Every template includes these sections:

Section	Purpose	Required?
Goal	One sentence — what done looks like	Yes
Context	Current state + problem (max 5 bullets)	Yes
Scope	In scope / Out of scope	Yes
Constraints	What must NOT change	Yes
Acceptance Criteria	Binary pass/fail checkboxes	Yes
Related	File paths, plan links	Yes
Agent State	Resumability checkpoint	Yes

Writing good acceptance criteria

Good (binary, verifiable):

• [ ] All existing tests pass after changes
• [ ] Login endpoint returns 401 for invalid credentials
• [ ] Response time < 200ms for P95

Bad (subjective, ambiguous):

• [ ] Code is clean — what does "clean" mean?
• [ ] Performance is good — how good?
• [ ] Handles edge cases — which ones?

File naming

YYMMDD-kebab-description.type.md

Examples:
260327-add-user-authentication.feature.md
260327-fix-login-timeout.bug-fix.md
260327-extract-auth-middleware.refactor.md

Plan Creator Reference System

The meow:plan-creator skill uses progressive disclosure — details live in reference files loaded on-demand:

Reference	Purpose	When loaded
scope-challenge.md	Scope modes (HOLD/EXPANSION/REDUCTION), complexity thresholds	Before research begins
research-phase.md	Scout + researcher subagent protocol	When research is needed
plan-organization.md	Directory structure, naming, size rules	When creating plan directory
output-standards.md	YAML frontmatter, required sections, quality rules	When drafting plan
solution-evaluation.md	Trade-off scoring for multiple options	When comparing approaches
task-management.md	Hydration, cross-session resume, sync-back	After plan approval
gotchas.md	Known planning failure modes	Always (inline top 3 in SKILL.md)
workflow-models/*.md	Phase flow per task type	After model selection

Plans validate via scripts/validate-plan.py — deterministic checks for required sections, binary acceptance criteria, and non-empty constraints.