Claude Code Plugins

Community-maintained marketplace

Feedback

change-maintenance-specification

@Z1-Test/temp-dhruvil-st15
0
0

Specifies bugs, upgrades, refactors, and behavioral changes as explicit deltas against existing feature specifications, preserving original intent while defining what changes, what remains unchanged, and how correctness is verified.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name change-maintenance-specification
description Specifies bugs, upgrades, refactors, and behavioral changes as explicit deltas against existing feature specifications, preserving original intent while defining what changes, what remains unchanged, and how correctness is verified.

Change & Maintenance Specification

Purpose

This skill produces a change specification for work that modifies existing behavior.

It is used when a feature, workflow, or system capability already exists and must be:

  • corrected (bug),
  • improved without changing behavior (upgrade/refactor), or
  • intentionally altered (behavioral change).

This skill does not replace existing feature specifications.
It defines a delta against them.


When to Use This Skill

Use this skill when:

  • A bug is reported in an existing feature
  • A refactor or upgrade is required without changing user-visible behavior
  • An intentional change modifies how an existing feature behaves
  • Regression risk must be explicitly controlled

Do not use this skill to:

  • define new features
  • rewrite PRDs
  • decompose features
  • create implementation plans or code
  • manage workflows or execution

Core Principle (Non-Negotiable)

Existing feature specifications define intended behavior.
Change specifications define deviations and corrections.

A bug or upgrade must not silently rewrite intent.


Supported Change Types

This skill supports all of the following change types.

1. Bug Fix

  • Current behavior is incorrect
  • Intended behavior already exists or is implied by the feature specification

2. Upgrade / Refactor

  • Behavior is correct
  • Internal structure, safety, performance, or maintainability changes
  • User-visible behavior must remain unchanged

3. Behavioral Change

  • Existing behavior is intentionally modified
  • User-visible outcomes may change
  • Change must be explicit and auditable

4. Feature Flag Lifecycle

  • Flag Removal: Removing a temporary flag once exit criteria are met.
  • Flag Promotion: Converting a temporary flag to a permanent one based on new justification.
  • Flag Behavior Modification: Changing how an existing flag controls logic.

Delta-Based Thinking

Every change specification must clearly distinguish:

  • Baseline — what exists today according to the feature specification (including current flag state)
  • Current Reality — what actually happens (if different)
  • Target Behavior — what should happen after the change (including flag removal or promotion)
  • Unchanged Behavior — what must remain exactly the same

Flag changes are treated as DELTAS. Do not allow silent flag changes.

This separation prevents regression and scope creep.


Change Specification Structure

The output of this skill must follow the structure below.

Change Specification

1. Change Summary

2. Baseline Reference

3. Current Behavior

4. Target / Expected Behavior

5. Scope of Change

6. Unchanged Behavior

7. Out of Scope

8. Impacted Areas

9. Risk & Regression Considerations

10. Scenarios

11. Acceptance Criteria

1. Change Summary

Briefly describe:

  • what is changing
  • why the change is required
  • the type of change (bug, upgrade, behavioral change)

2. Baseline Reference

Explicitly reference the existing feature specification or documentation that defines the intended behavior.

This establishes the contract against which the change is evaluated.

3. Current Behavior

Describe:

  • what actually happens today
  • when the issue occurs
  • observable incorrect or undesirable outcomes

Be factual and neutral.

4. Target / Expected Behavior

Describe:

  • the behavior that should exist after the change
  • alignment with the baseline specification (for bugs/upgrades)
  • intentional differences (for behavioral changes)

Avoid proposing implementation details.

5. Scope of Change

Clearly state:

  • what behavior is being modified
  • what components or flows are affected conceptually

This defines the blast radius.

6. Out of Scope

Explicitly state what this change does not affect.

This is critical for preventing accidental expansion.

7. Impacted Areas

List:

  • features
  • workflows
  • data paths
  • user experiences that could be affected by the change.

This is informational, not a task list.

8. Risk & Regression Considerations

Identify:

  • potential regressions
  • user trust risks
  • data consistency risks
  • compatibility concerns

Do not include mitigation plans—only awareness.

9. Scenarios

Provide scenarios that illustrate:

  • the problematic behavior (before)
  • the corrected or changed behavior (after)
  • confirmation that unchanged behavior remains intact

Scenarios may use Gherkin format when helpful:

### Scenario: [Description]
Given ...
When ...
Then ...

Scenarios should emphasize contrast, not completeness.

10. Acceptance Criteria

Define objective conditions that confirm the change is complete.

Acceptance criteria must:

  • be verifiable
  • map to scenarios
  • include regression confirmation where relevant

Avoid vague or subjective language.


Important Boundaries

This skill must not:

  • rewrite feature specifications wholesale
  • propose implementation approaches
  • assign tasks, owners, or timelines
  • decide rollout or deployment strategy
  • create or modify files or issues
  • control execution or workflow state

All orchestration belongs to the calling agent.


Output Expectations

The output of this skill should be:

  • concise but precise
  • explicitly delta-focused
  • auditable over time
  • safe to use in long-lived systems
  • deterministic for the same inputs

Assume the output will be reviewed by engineers, reviewers, and future maintainers to understand what changed and why.