docs: additive-vs-reductive test change detection design spec (TC-4153)

[mrizzi]

Summary

• Design spec for adding test change classification to verify-pr's Step 12 (Test Quality)
• Distinguishes additive test changes (new coverage) from reductive ones (weakened coverage) using a sub-agent with hybrid structural scan + semantic assessment
• Validated against two real trustify PRs: #2325 (purely additive) and #2318 (mixed — predominantly additive with one relaxed assertion)

Closes spike TC-4153

Test plan

• Spec self-review: no placeholders, no contradictions, no ambiguity, scoped for single implementation plan
• Validated detection approach against real PR diffs
• User reviews spec before implementation planning

🤖 Generated with Claude Code

Summary by Sourcery

Documentation:

• Add a detailed design spec describing additive vs reductive test change classification for verify-pr, including architecture, classification logic, reporting, and SKILL.md integration.

[sourcery-ai]

Reviewer's Guide

Adds a detailed design spec document describing how verify-pr will classify test changes as additive vs reductive using a dedicated sub-agent that performs a hybrid structural scan and semantic assessment, integrates its results into Step 12 (Test Quality) and Step 14 reporting, and outlines SKILL.md updates and future autonomous-mode escalation behavior.

File-Level Changes

Change	Details	Files
Document sub-agent architecture for additive vs reductive test change classification in verify-pr	Describe main verify-pr agent responsibilities for detecting test file changes and invoking a sub-agent Specify sub-agent inputs (modified/deleted test paths and branch names) and outputs (classification, structural summary, semantic assessment, reductive findings) Define how the main agent consumes sub-agent results without re-reading test contents	docs/specs/2026-04-16-additive-vs-reductive-detection-design.md
Define hybrid structural scan and semantic assessment approach for evaluating test changes	List structural signals to count per test file, including test functions, assertions, skip annotations, parameterized cases, and mock scope changes Describe language-agnostic but framework-aware detection of assertions and related patterns Detail semantic assessment process for comparing behaviors under test and using it to override misleading structural signals	docs/specs/2026-04-16-additive-vs-reductive-detection-design.md
Specify classification logic and how it affects verify-pr reporting and outcomes	Formalize criteria for ADDITIVE, REDUCTIVE, MIXED, NEUTRAL, and N/A classifications and the role of semantic assessment as tiebreaker Describe how overall classification combines sub-agent output with new-file analysis Outline how the new Test Change Classification row integrates into the Step 14 verification report and its advisory-only effect	docs/specs/2026-04-16-additive-vs-reductive-detection-design.md
Outline SKILL.md integration and step sequencing for the new test change classification behavior	Describe where in verify-pr/SKILL.md the new sub-steps land and how existing Step 12 sub-steps are renumbered Define new sub-steps for identifying test file change types, spawning the sub-agent, running structural/semantic analysis, and recording findings Clarify that no new runtime behavior is implemented yet; this document only specifies planned SKILL.md changes	docs/specs/2026-04-16-additive-vs-reductive-detection-design.md
Provide concrete real-world examples validating the design	Include a purely additive example PR demonstrating ADDITIVE classification and the expected report row Include a mixed-change example PR showing MIXED classification, with one relaxed assertion and summarized reductive finding in the report row	docs/specs/2026-04-16-additive-vs-reductive-detection-design.md

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[sourcery-ai]

Hey - I've left some high level feedback:

• The Step 12 sub-step renumbering in the SKILL.md integration section is a bit confusing (existing steps 1–7 vs. new steps 8–13); it would help to explicitly map old → new step numbers so someone editing SKILL.md can apply the changes without ambiguity.
• In the structural scan section, consider clarifying how the sub-agent identifies the test framework / language for a file (e.g., by extension, heuristics, or upstream metadata) and what it should do when it cannot confidently recognize assertions or skip annotations.
• For the sub-agent interface, it might be worth specifying a more structured output format (e.g., JSON schema for classification, structural summary, and reductive findings) so that the main agent can consume the results reliably and avoid parsing free-form text.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- The Step 12 sub-step renumbering in the SKILL.md integration section is a bit confusing (existing steps 1–7 vs. new steps 8–13); it would help to explicitly map old → new step numbers so someone editing SKILL.md can apply the changes without ambiguity.
- In the structural scan section, consider clarifying how the sub-agent identifies the test framework / language for a file (e.g., by extension, heuristics, or upstream metadata) and what it should do when it cannot confidently recognize assertions or skip annotations.
- For the sub-agent interface, it might be worth specifying a more structured output format (e.g., JSON schema for classification, structural summary, and reductive findings) so that the main agent can consume the results reliably and avoid parsing free-form text.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}