C/C++test AI Agent Skills

Table of Contents

• Overview
• Prerequisites
• Skills overview
• Integrating skills into your project
• Basic configuration
• Additional configuration for GitHub CI/CD
• Using skills with C/C++test Professional
• Using skills with Windows PowerShell

Overview

This repository contains C/C++test static analysis skills for AI coding agents, enabling you to fix static analysis violations both autonomously in CI/CD and interactively in a local environment.

A pre-configured demo project is available at https://github.com/parasoft/cpptest-ai-agent-demo.

Prerequisites

• C/C++test installed and configured
• AI coding agent installed and configured
• GitHub CLI installed - for using the GitHub CI/CD pipeline

Hint: An exemplary Dockerfile is available in the devcontainer folder. See devcontainer/README.md for build and usage details.

Skills overview

1. cpptest-fix-all-violations — Runs C/C++test static analysis and iteratively fixes all reported violations, rule by rule. After fixing each rule, the fixes are verified. Supports configurable iteration and rule limits, and can optionally commit fixes after each rule. Analysis is performed by executing the cpptest-analyze.sh script.

   How to use: the primary use case is autonomous fixing inside a CI/CD pipeline (see Additional configuration for GitHub CI/CD below). The skill can also be used locally by prompting the AI agent, e.g. "Fix all violations in this project" or "Fix all MISRA violations". To fix a single, already-known violation, use cpptest-fix-one-violation instead.

   Configuration environment variables:

   • CPPTEST_COMMIT_FIXES — If set to true or 1, commit fixes after each rule. Default: not set.
   • CPPTEST_ITERATION_LIMIT — Maximum number of fix-and-verify attempts per rule before declaring it unfixable. Default: 3.
   • CPPTEST_RULES_LIMIT — Maximum number of rules to process before stopping. Default: 20.

2. cpptest-fix-one-violation — Fixes a single, already-known C/C++test static analysis violation described by the user. Does not run analysis upfront — violation details are taken directly from the user's prompt. Verification is only performed when explicitly requested, and is performed by executing the cpptest-analyze.sh script.

   How to use: prompt the AI agent with the details of a specific violation, e.g. "Fix and verify CERT_C-DCL00-a: Declare local variable 'sensor' as const at sensor.c:55". To fix multiple or all violations, use cpptest-fix-all-violations instead.

Integrating skills into your project

Assuming the project is located at <PROJECT_DIR>.

1. Copy the skills folder into your project — typically into <PROJECT_DIR>/.agents. See the agent documentation for the location of skill definition files.
2. Copy the agents/AGENTS.md file into <PROJECT_DIR>. See the agent documentation for the location of agent instruction files.
3. Copy the scripts/cpptest-analyze.sh file into <PROJECT_DIR>.
4. To use the GitHub Actions pipeline, copy the workflows folder into <PROJECT_DIR>/.github.

A typical project layout looks like this:

<PROJECT_DIR>
  .agents/skills/
     cpptest-fix-all-violations/SKILL.md
     cpptest-fix-one-violation/SKILL.md
  .github/workflows/
     cpptest-agent-prompt.md
     cpptest-agent-run.sh
     cpptest-autofix-github.yml
  AGENTS.md
  cpptest-analyze.sh

Basic configuration

Review and update the following configuration files:

• cpptest-analyze.sh should run the C/C++test static analysis with additional build and verification steps. It is intended to be used by the C/C++test skills, whenever the static analysis needs to be re-executed or applied fixes need to be verified. The skills expect the static analysis report file to be created in <PROJECT_DIR>/reports/report.xml; adjust the skill definition files (SKILL.md) if needed.

Additional configuration for GitHub CI/CD

Review and update the following configuration files in .github/workflows:

• cpptest-autofix-github.yml should execute the build pipeline with C/C++test static analysis and automatically fix the reported static analysis violations. Before fixing the violations, a separate branch is created (autofix/pr-<number>/<timestamp>). After violations are fixed, the changes are committed (via CPPTEST_COMMIT_FIXES=1), the autofix branch is pushed to the remote repository, and a pull request is created automatically.
• cpptest-agent-run.sh should configure and run an AI agent with the prompt defined in cpptest-agent-prompt.md. Adjust to your specific AI coding agent.

See the Run C/C++test AI Autofix step in cpptest-autofix-github.yml for details.

Note: To allow GitHub Actions to create pull requests automatically, enable the following option in your repository: Settings > Actions > General > Allow GitHub Actions to create and approve pull requests.

Using skills with C/C++test Professional

The integration scripts and workflow in this project are configured for C/C++test Standard. To use them with C/C++test Professional, review and update both the local analysis command (cpptest-analyze.sh) and the GitHub workflow configuration (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"

If you use the GitHub CI/CD pipeline, review the Run C/C++test step in 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.

Using skills with Windows PowerShell

By default, the skills are prepared to work with a bash shell. To use the skills in a Windows-based environment with PowerShell, the cpptest-analyze.sh and cpptest-agent-run.sh scripts must be converted to PowerShell scripts (cpptest-analyze.ps1 and cpptest-agent-run.ps1). Then update the SKILL.md files: replace cpptest-analyze.sh references with cpptest-analyze.ps1, and change the script invocation code block language from bash to powershell.

Additionally, for the GitHub CI/CD pipeline, review and adjust cpptest-autofix-github.yml (replace cpptest-agent-run.sh with its .ps1 counterpart).