Thank you for your interest in improving test quality governance.
-
Define the violation in
references/violation-types.md:- Name (SCREAMING_SNAKE_CASE)
- Category (A: Assertion Weakness, B: Test Design, C: Coverage Gaps, D: Test Integrity)
- Default severity (critical, error, warning, info)
- Detection method
- Examples in at least 2 languages
- "Why It Matters" explanation
-
Add to the protocol in
templates/litmus-protocol.yaml:- Add the violation name under
violations:with its default severity - Add penalty points under
penalties:if introducing a new severity
- Add the violation name under
-
Update the test-audit skill in
.claude/skills/test-audit/SKILL.md:- Add detection logic under the appropriate category step
- Include in the verdict calculation
-
Update the assertion-analyzer agent in
.claude/agents/assertion-analyzer.md:- Add detection rules
-
Update adapters:
adapters/claude-code/commands/litmus-scan.mdadapters/codex/AGENTS.mdadapters/cursor/.cursor/rules/litmus.mdadapters/aider/.aider.conf.yml
-
Update prompts:
prompts/litmus-scan.md
-
Update
references/assertion-classification.md:- Add a new section for the framework
- Provide STRONG, WEAK, and HOLLOW examples
- Note any framework-specific edge cases
-
Update the assertion-analyzer agent if the framework uses unusual assertion syntax.
-
Create the adapter directory under
adapters/{cli-name}/ -
Write the configuration file following the target CLI's conventions:
- System prompt integration
- Rule files
- Configuration files
-
Update the installer in
install.sh:- Add detection logic in
detect_cli() - Add setup logic in the
caseblock
- Add detection logic in
-
Update
adapters/generic/README.mdwith the new platform in the support table.
-
Update
references/edge-case-taxonomy.md:- Add the category with description
- Include concrete examples
- Add to the priority matrix
-
Update the edge-detector agent in
.claude/agents/edge-detector.md. -
Update the edge-analysis skill in
.claude/skills/edge-analysis/SKILL.md. -
Update
templates/litmus-protocol.yamlunderedge.categories.
- Skills use step-by-step numbered workflows
- Agents use role + capabilities + rules format
- Commands use YAML frontmatter + invocation instruction
- References use tables + code examples
- All files are Markdown (except YAML configs and JSON metadata)
Since Agent-Litmus is a prompt-based plugin (no runtime code), testing means:
- Run each modified command against a sample test file
- Verify the output matches the expected format
- Check that new violation types are detected correctly
- Verify scoring changes produce reasonable TQS values
- Fork the repository
- Create a feature branch:
feature/add-violation-type-x - Make your changes across all affected files
- Submit a PR with:
- What violation/feature you added
- Example detection output
- Which files were modified
By contributing, you agree that your contributions will be licensed under the MIT License.