Skip to main content
Version: 1.x

Plugin Commands Reference

In this guide
  • Run /praman-plan to explore a live SAP app and produce a structured test plan
  • Run /praman-generate to convert a plan into compliant Praman .spec.ts files
  • Run /praman-heal to automatically diagnose and fix failing tests with confidence scoring
  • Run /praman-coverage for the full pipeline: plan, generate, and heal with user gates
  • Understand the agents dispatched, artifacts produced, and user gates at each phase

The Praman Claude Code plugin provides four slash commands that form a complete SAP test automation pipeline. Each command dispatches specialized AI agents, produces specific artifacts, and pauses at user gates for review before proceeding.

Pipeline Overview​

flowchart LR
    subgraph "/praman-coverage (wraps all three)"
        A["/praman-plan"] -->|specs/*.plan.md| G1{User Gate}
        G1 -->|approve| B["/praman-generate"]
        B -->|tests/*.spec.ts| G2{User Gate}
        G2 -->|approve| C["/praman-heal"]
        C -->|fixed *.spec.ts| G3{User Gate}
        G3 -->|approve| D["All tests passing"]
    end

Command Reference​

/praman-plan​

Purpose: Explore a live SAP Fiori app and produce a structured test plan.

Argument hint: A natural-language description of the app and scenarios to cover.

Example invocation:

/praman-plan
"Explore the Manage Purchase Orders app and plan tests for creating
a PO with 2 line items, editing quantity, and deleting a line item"

What happens step-by-step:

  1. Dispatches sap-explorer agent to open the SAP app in a browser session
  2. sap-explorer authenticates, navigates FLP, and discovers all UI5 controls
  3. Dispatches sap-architect agent to design test scenarios from the discovery data
  4. sap-architect structures scenarios with control IDs, types, and binding paths
  5. Writes the test plan to specs/<app-name>.plan.md
  6. User gate: Presents the plan for review before proceeding

Output artifacts:

ArtifactLocationContents
Test planspecs/<app>.plan.mdScenarios, steps, control selectors, data values
Discovery snapshotsspecs/snapshots/*.ymlPage structure at each navigation point
Gold-standard referencespecs/<app>.gold.mdExpected fixture calls for each step

/praman-generate​

Purpose: Load an existing plan and generate compliant Praman test files.

Argument hint: Path to the plan file, or omit to auto-detect the most recent plan.

Example invocations:

/praman-generate specs/manage-purchase-orders.plan.md
/praman-generate

What happens step-by-step:

  1. Loads the plan file and parses scenarios
  2. Dispatches test-generator agent for each scenario in the plan
  3. test-generator writes .spec.ts files using Praman fixtures exclusively
  4. Runs the quality gate: 19 forbidden patterns + 7 mandatory rules
  5. If violations found, auto-fixes up to 3 iterations
  6. Dispatches code-reviewer agent for final compliance check
  7. User gate: Presents generated files for review

Output artifacts:

ArtifactLocationContents
Test filestests/e2e/<app>.spec.tsPraman-compliant test code
Quality reportspecs/<app>.quality.mdRule violations found and auto-fixed
Review commentsInline in conversationcode-reviewer findings and suggestions

Quality gate details:

The 19 forbidden patterns include page.click('#__...'), page.waitForTimeout(), @playwright/test imports, and raw DOM selectors. The 7 mandatory rules enforce playwright-praman imports, test.step() structure, setValue() + fireChange() + waitForUI5() sequences, and TSDoc headers. Auto-fix attempts up to 3 iterations before presenting remaining issues to the user.

/praman-heal​

Purpose: Run failing tests, diagnose failures, and apply fixes automatically.

Argument hint: Path to a specific test file, or omit to heal all failing tests.

Example invocations:

/praman-heal tests/e2e/manage-purchase-orders.spec.ts
/praman-heal

What happens step-by-step:

  1. Runs the test suite and collects failure results
  2. Dispatches test-healer agent to classify each failure
  3. test-healer assigns a confidence score (0-100) to each proposed fix
  4. Fixes with confidence >= 80 are auto-applied without user confirmation
  5. Fixes with confidence < 80 are presented for user review
  6. Re-runs tests after applying fixes (max 3 healing iterations)
  7. User gate: Presents final results and any remaining failures

Confidence scoring:

Score RangeBehaviorExample
80-100Auto-appliedSelector ID changed, timeout too short
50-79Suggested, user reviewLogic change, different control type
0-49Reported onlyApp behavior changed, new dialog

Output artifacts:

ArtifactLocationContents
Fixed test filetests/e2e/<app>.spec.tsUpdated test code with fixes applied
Healing reportspecs/<app>.healing.mdFailures classified, fixes applied
Trace filestest-results/Playwright traces for failed runs

/praman-coverage​

Purpose: Run the full pipeline -- plan, generate, and heal -- with user gates between each phase.

Argument hint: A natural-language description of the app and desired test coverage.

Example invocation:

/praman-coverage
"Generate complete test coverage for the Create Purchase Order app
including happy path, validation errors, and draft handling"

What happens step-by-step:

  1. Runs /praman-plan (dispatches sap-explorer + sap-architect)
  2. User gate 1: Review and approve the test plan
  3. Runs /praman-generate (dispatches test-generator + code-reviewer)
  4. User gate 2: Review and approve generated test files
  5. Runs /praman-heal (dispatches test-healer)
  6. User gate 3: Review final results and any remaining failures
  7. Reports overall coverage summary

Command Comparison​

Capability/praman-plan/praman-generate/praman-heal/praman-coverage
Needs live SAP systemYesNoYesYes
Needs existing planNoYesNoNo
Needs existing testsNoNoYesNo
Agents dispatched2215
User gates1113

Agents dispatched per command:

Commandsap-explorersap-architecttest-generatortest-healercode-reviewer
/praman-planYesYes------
/praman-generate----Yes--Yes
/praman-heal------Yes--
/praman-coverageYesYesYesYesYes
Common mistake

Running /praman-generate without a plan file. The generate command requires an existing specs/*.plan.md file produced by /praman-plan. If no plan file exists, the generator will fail immediately. Always run /praman-plan first, or use /praman-coverage to run the full pipeline automatically.

Common mistake

Expecting the pipeline to run unattended. Every command includes at least one user gate where the pipeline pauses and waits for your approval. /praman-coverage has three user gates (after plan, after generate, after heal). This is by design -- the gates let you review and correct agent output before the next phase uses it as input. Do not assume the pipeline runs start-to-finish without interaction.

FAQ​

Can I skip the plan phase and go straight to generating tests?

Yes, if you already have a plan file. Run /praman-generate specs/my-app.plan.md directly. However, the plan file must follow the expected format with scenarios, steps, and control selectors. Manually authored plans work as long as they include the required sections. Run /praman-plan at least once to see the expected format, then use it as a template.

How do I re-run healing on a single test file?

Pass the file path directly to the heal command:

/praman-heal tests/e2e/manage-purchase-orders.spec.ts

This runs only that file, classifies its failures, and attempts fixes. The healer will not touch other test files in the project.

What happens if I reject the plan at the user gate?

The pipeline pauses and you can provide feedback. Tell the agent what to change -- for example, "add a scenario for validation errors" or "remove the draft handling scenario." The planner will revise the plan and present it again for approval. You can reject and revise as many times as needed before proceeding to generation.

Can I run commands on a non-SAP Playwright app?

No. The Praman commands are designed specifically for SAP UI5 and Fiori applications. They rely on the UI5 control tree, sap.ui.core.ElementRegistry, and Praman bridge injection. For non-SAP apps, use standard Playwright test authoring instead.

Next steps