agentsclimarketplace

Recipe add integration tests

Skill shinpr/claude-code-workflows/skills/recipe-add-integration-tests

Add integration/E2E tests to existing codebase using Design DocsFrom its SKILL.md

Install
npx -y skills add shinpr/claude-code-workflows --skill recipe-add-integration-tests

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

8.2 KB, as published. Nobody here has run it

Explicit User Instruction: The user explicitly instructs and authorizes every subagent call named in this recipe. Execute each applicable call when its prerequisites are met.

Execute Skill: llm-friendly-context before writing Agent prompts, handoffs, or generated artifacts. Execute Skill: subagents-orchestration-guide before making workflow decisions, invoking agents, or resolving findings.

Context: Test addition workflow for existing implementations (backend, frontend, or fullstack)

Orchestrator Definition

Core Identity: "I am an orchestrator."

Local authority gate: Make this recipe's workflow decisions and validate each returned result directly; delegate semantic deliverable production to the named specialist.

Review Resolution Gate [MANDATORY]: Resolve every actionable deliverable-review finding through subagents-orchestration-guide Review Resolution before correction or progression. Before the first finding disposition, read references/review-resolution.md from the loaded subagents-orchestration-guide skill.

Execution Gate: Complete Steps 1-7 in order for each generated layer. Advance only through the current step's stated output or response gate; skip work only when its stated condition is false. Report completion after every layer has completed its review, quality, commit, and retained-limitation retry.

Why Delegate: Orchestrator's context is shared across all steps. Direct implementation consumes context needed for review and quality check phases. Subagents work in isolated context.

Execution Method:

  • Skeleton generation → delegate to acceptance-test-generator
  • Test implementation → delegate to task-executor
  • Test review → delegate to integration-test-reviewer
  • Quality checks → delegate to quality-fixer

At each Agent invocation below, build the prompt as a mechanical extraction: copy the named source values into the exact fields, apply only the declared serialization, then invoke immediately.

Document paths: $ARGUMENTS

Prerequisites

  • At least one Design Doc must exist (created manually or via reverse-engineer)
  • Existing implementation to test

Execution Flow

Step 1: Discover and Validate Documents

Confirm $ARGUMENTS names at least one existing document path; report and end when it is empty or every path is unresolvable. Then discover the remaining Design Docs, UI Specs, and PRDs under docs/design/, docs/ui-spec/, and docs/prd/.

Classify discovered documents by filename:

  • Filename contains backend → Design Doc (backend)
  • Filename contains frontend → Design Doc (frontend)
  • Located in docs/ui-spec/ → UI Spec (optional)
  • Located in docs/prd/ → approved PRD / confirmed requirement context (optional; prefer the path named by the Design Doc)
  • None of the above → treat as single-layer Design Doc

Step 2: Skeleton Generation

Invoke acceptance-test-generator using Agent tool:

  • subagent_type: "dev-workflows:acceptance-test-generator"
  • description: "Generate test skeletons"
  • design_docs: Existing backend, frontend, or single-layer Design Doc paths from Step 1
  • ui_spec: Existing UI Spec path when present
  • confirmed_requirement_context: Approved PRD path or unchanged Design Doc Requirement Convergence record

Follow subagents-orchestration-guide HC-06 for value_input_required and its unknown-value continuation before Step 3.

Expected output: generatedFiles containing integration and e2e paths

Step 3: Test Implementation

For each layer with generated skeletons, record the current HEAD as diffBase, then invoke the layer's task-executor:

  • Backend or single-layer → subagent_type: "dev-workflows:task-executor"
  • Frontend → subagent_type: "dev-workflows-frontend:task-executor-frontend"
  • description: "Implement integration tests"
  • direct_scope: Implement every test defined by the layer-specific generated skeletons
  • governing_sources: Layer-specific Design Doc, applicable UI Spec, and generated skeleton paths
  • target_paths: Generated test paths plus the existing setup or fixture paths explicitly identified before invocation
  • observable_verification: Execute the implemented tests and verify every skeleton claim at its declared boundary

Execute one layer at a time through Steps 3→4→5→6→7 before starting the next.

Expected output: status, testsAdded, mutationEvidence

Apply this response gate after every task-executor invocation in Steps 3 and 5:

  • At least one changed integration/E2E test file and its implementation evidence are confirmed from the response and repository state → Proceed to Step 4
  • Required implementation remains incomplete → Apply Specialist Result Acceptance and continue Step 3 while repository evidence supplies an advancing action
  • A changed product outcome, major approved design change, user-held authority, or irreversible action is identified → Present that decision to the user
  • Other result variations → Apply subagents-orchestration-guide Specialist Result Acceptance

Step 4: Test Review

Invoke integration-test-reviewer using Agent tool:

  • subagent_type: "dev-workflows:integration-test-reviewer"
  • description: "Review test quality"
  • prompt: "Review test quality. changedTestFiles: [confirmed changed integration/E2E test paths]. diffBase: [revision recorded before Step 3]. skeletonFiles: [layer-specific paths from Step 2 generatedFiles]. mutationEvidence: [Step 3 mutationEvidence]."

Expected output: status (approved/needs_revision/blocked), testFiles, reviewBasis, qualityIssues; correction re-review also returns prior_feedback_reconciliation

Step 5: Apply Review Fixes

Check Step 4 result:

  • status: approved → Mark complete, proceed to Step 6
  • status: blocked → Apply Specialist Result Acceptance
  • status: needs_revision → Pass Step 4 qualityIssues unchanged into the Review Resolution Gate; invoke task-executor for rerouted corrections, return to Step 4, and derive convergence from prior_feedback_reconciliation

Invoke the same layer's task-executor:

  • description: "Fix review findings"
  • Reuse Step 3 direct_scope, governing_sources, target_paths, and observable_verification
  • correction_findings: Complete reviewer finding objects verbatim, with only their orchestrator dispositions added

Step 6: Quality Check

Invoke quality-fixer for the current layer:

  • Backend or single-layer → subagent_type: "dev-workflows:quality-fixer"
  • Frontend → subagent_type: "dev-workflows-frontend:quality-fixer-frontend"
  • description: "Final quality assurance"
  • Pass the latest executor's mutationEvidence.
  • prompt: "Run the repository-configured quality checks applicable to the test files added in this workflow and verify their intended observable behavior."

Expected output: status (approved, stub_detected, verification_incomplete, or blocked)

Check quality-fixer response:

  • stub_detected → Return to Step 3 with the quality-fixer's incompleteImplementations array unchanged as the canonical incompleteImplementations field, then re-execute Steps 3→4→5→6
  • blocked → Apply Specialist Result Acceptance
  • verification_incomplete → Retain the complete result for one final retry and proceed to Step 7
  • approved → Proceed to Step 7

Step 7: Commit

On approved or verification_incomplete from quality-fixer:

  • Apply subagents-orchestration-guide Commit Boundary Check, then commit test files using Bash with message format: "test: add [layer] integration tests for [feature name]". Append its verification trailers for verification_incomplete.

After every layer has a clean commit boundary, retry each retained verification limitation once with the same layer quality-fixer inputs. Clear an approved result, route newly discovered incomplete implementation through Steps 3→6, and retain a repeated limitation for the completion report while continuing the workflow.

In the completion report, list each repeated verification limitation and each declined actionable finding with its ID, governing reason, and evidence when any occurred.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most test skills give

Counted across 1,201 of the 2,096 authors here whose files we hold, read 2026-09-06

  • Write a failing test before writing codein 43 of 1201, across 36 files
  • Run the full test suitein 36 of 1201, across 35 files
  • Test only one variable per experimentin 34 of 1201, across 17 files
  • Read product marketing context before asking questionsin 34 of 1201, across 14 files
  • Mock external dependenciesin 34 of 1201, across 30 files
  • Define primary, secondary, and guardrail metricsin 33 of 1201, across 16 files
  • Pre-determine sample size before startingin 31 of 1201, across 14 files
  • Test behavior rather than implementationin 31 of 1201, across 29 files
  • Formulate a hypothesis before designing a testin 30 of 1201, across 13 files
  • Document every test hypothesis, variant, and resultin 29 of 1201, across 11 files
  • Use descriptive test function namesin 25 of 1201, across 21 files
  • Commit to the methodology without stopping earlyin 24 of 1201, across 8 files

Said here and by no other author read

  • Validate document paths before starting
  • Classify documents by filename and location
  • Execute steps one through seven in order
  • Delegate semantic production to named specialists
  • Resolve actionable findings via orchestration guide
  • Build prompts by mechanical extraction of source values

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.

Keep looking

Skills are one crate of 613,297. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.