domains / security / vulnerability-remediation / fix

fix

Skill card

combined

The two skills share the name 'fix' but target completely different tasks. The first is a security skill that fetches fix intelligence for a vulnerability and proposes remediation for the current repository, with strict reporting requirements for version provenance and confidence scores. The second is a software-engineering skill for diagnosing and fixing failing or flaky Playwright tests using a specific taxonomy. They do not overlap in steps, constraints, or intent, so crowning one would hide a real alternative.

Repositories

Other copies

Combined skill

---
name: fix
description: Fix vulnerabilities or failing Playwright tests based on context
---

# Fix Skill

This skill handles two distinct remediation tasks. Determine which applies based on the user's request and context.

## Scenario A: Vulnerability Remediation

Use this path when the user provides a vulnerability ID or asks for fix intelligence for a security issue.

1. **Fetch Fix Intelligence**: Retrieve details for the vulnerability.
2. **Analyze Repository**: Identify the affected package and its current version.
   - Determine version source (Manifest, Lockfile, Installed, etc.).
   - Search filesystem if version is not in context.
3. **Propose Remediation**:
   - Identify fix target version and source.
   - Calculate Safe Harbour Confidence (0.00–1.00).
   - Present options using Markdown and Mermaid diagrams for upgrade paths.
4. **Report**:
   - Include Current Version, Version Source, Fix Target Version, Fix Source, and Safe Harbour Confidence.
   - Use `jq`/`yq` for data processing if available.

## Scenario B: Playwright Test Fixing

Use this path when the user reports a failing or flaky Playwright test.

1. **Reproduce**: Run the test to capture the error. If it passes, run burn-in (`--repeat-each=10`).
2. **Capture Trace**: Run with `--trace=on`.
3. **Categorize**: Classify the failure as Timing/Async, Test Isolation, Environment, or Infrastructure.
4. **Apply Fix**:
   - **Timing/Async**: Replace `waitForTimeout` with web-first assertions.
   - **Test Isolation**: Remove shared state, use unique IDs.
   - **Environment**: Match viewport/fonts, check CI vs local diffs.
   - **Infrastructure**: Adjust timeouts, retries, or browser dependencies.
5. **Verify**: Run the test 10 times to confirm stability.
6. **Prevent**: Suggest CI retries and trace configuration.

## Conflicts

- **davepoon/buildwithclaude**: Focuses on security vulnerability remediation with strict version provenance reporting.
- **LeoYeAI/openclaw-master-skills**: Focuses on Playwright test debugging with a specific failure taxonomy.

These are mutually exclusive workflows. Select the appropriate scenario based on the input arguments and context.