C/C++test AI Agent Skills Demo

Table of Contents

• Overview
• Prerequisites
• Workflow
• Project layout
• Using project with C/C++test Professional

Overview

This repository is an example C++ project that demonstrates how the Parasoft C/C++test AI agent skills can be used to automatically fix static analysis violations inside a GitHub CI/CD pipeline.

The key idea is that when a pull request is created, the pipeline runs C/C++test, detects violations, and then invokes an AI coding agent (OpenAI Codex, or others) equipped with C/C++test skills and MCP tools to automatically fix the violations, commit the fixes to a new branch, and open a new pull request.

The C/C++test AI Agent skills used in this project are taken from the parasoft/cpptest-ai-agent-skills repository.

Prerequisites

The CI/CD pipeline requires a GitHub self-hosted runner configured and connected to the repository. The runner must have the following components available:

• Parasoft C/C++test Standard — installed, configured, and licensed
  • Expected installation path: /opt/parasoft/cpptest
  • The following directories should be on $PATH so that cpptestcli and cpptesttrace are accessible to the pipeline:
    • /opt/parasoft/cpptest
    • /opt/parasoft/cpptest/bin
• GitHub CLI - installed
• AI coding agent — OpenAI Codex CLI, authorized and ready to use
• C/C++ toolchain — GNU GCC compiler and make build tool

Note: If your C/C++test installation path or AI agent setup differs from the defaults, update the configuration in .github/workflows/cpptest-autofix-github.yml, .github/workflows/cpptest-agent-run.sh, and cpptest-analyze.sh accordingly.

Tip: Consider using a Docker container as your GitHub self-hosted runner with C/C++test and the AI agent pre-installed. An example Dockerfile is available in parasoft/cpptest-ai-agent-skills/devcontainer.

Workflow

1. Preparation

1. Fork or copy this repository.
2. Set up and connect a GitHub self-hosted runner: Settings > Actions > Runners. See also Prerequisites above.
3. Enable the following option in your repository: Settings > Actions > General > Allow GitHub Actions to create and approve pull requests.
4. Clone the repository to your local machine.

2. Introducing a Code Change

1. Create a feature branch in the local repository.
2. Introduce a change to the source code.
3. Push the branch to the remote GitHub repository.
4. Open a pull request on GitHub to merge your feature branch into your default branch (master or main).

3. CI/CD Pipeline Execution

Opening a pull request triggers the workflow defined in .github/workflows/cpptest-autofix-github.yml:

1. The project is built.
2. C/C++test runs static analysis and archives the reports as build artifacts.
3. If violations are found, the AI Autofix step is triggered:
   • A new autofix/pr-<number>/<timestamp> branch is created.
   • The AI agent processes violations rule by rule.
   • Each fix is verified by running cpptest-analyze.sh.
   • Verified fixes are committed to the autofix branch (one commit per rule).
   • The autofix branch is pushed to the remote repository.
   • A pull request from autofix into your feature branch is opened, and a comment with links is posted on the original PR.

4. Reviewing the Autofix Pull Request

1. Review the changes proposed in the autofix branch.
2. Merge the autofix branch into your feature branch.
3. Verify that the pipeline passes on the updated feature branch.
4. Merge the feature branch into your default branch (master or main).

Project layout

The repository is structured as follows, highlighting the key files related to C/C++test AI agent skills configuration:

<PROJECT_DIR>
  .agents/skills/
    cpptest-fix-all-violations/
      SKILL.md                    # AI agent skill: fix all violations in the project
    cpptest-fix-one-violation/
      SKILL.md                    # AI agent skill: fix a single known violation
  .github/workflows/
    cpptest-agent-prompt.md       # Prompt passed to the AI agent
    cpptest-agent-run.sh          # Script that invokes the AI agent
    cpptest-autofix-github.yml    # GitHub build pipeline
  include/
    *.hxx 
  *.cxx
  AGENTS.md                       # Agent instructions for this repository
  cpptest-analyze.sh              # Script for running C/C++test analysis
  Makefile

Using project with C/C++test Professional

This project is configured for C/C++test Standard. To use it with C/C++test Professional, review and update both the local analysis command (cpptest-analyze.sh) and the GitHub workflow configuration (.github/workflows/cpptest-autofix-github.yml) so they match your installation and project layout.

Update cpptest-analyze.sh so its cpptestcli command invocation matches your C/C++test Professional setup. In particular, verify the workspace location, test configuration, report directory, and analysis input file. For example:

cpptestcli -data ../workspace -config "builtin://Recommended Rules" -report "reports" -bdf "cpptestscan.bdf"

Review the Run C/C++test step in .github/workflows/cpptest-autofix-github.yml and adjust the run-cpptest-action inputs for C/C++test Professional. This typically includes settings such as the analysis input, workspace location, and any additional parameters required by your environment. See Customizing the Action to Run C/C++test Professional for details.