OH

open-harness-engineering/ai-agents-oss-helper

Developer tools
26 stars Quality 40 Trend 32

A collection of skills for AI coding agents (Claude, Bob, Gemini, OpenCode, Codex) that provides guidelines for contributing to open source projects.

Overview

A collection of skills for AI coding agents (Claude, Bob, Gemini, OpenCode, Codex) that provides guidelines for contributing to open source projects. The skills auto-detect the current project via git remote get-url origin and load project-specific configuration from rule files. For agents that support skills natively (Claude, Bob, Codex), the helper is installed as 6 focused, user-invocable skills — just describe what you want to do (or invoke /oss-* directly) and the agent follows the appropriate guidelines automatically. For other agents (Gemini, OpenCode), individual commands are generated from the same guidelines. Any project can use the helper by adding an .oss-ai-helper-rules/ directory to its repository root with three rule files: Use the Add Project guideline to generate initial rule files for any project. Once committed, every contributor gets the right configuration automatically — no per-user installation of project-specific rules required.

README

AI Agents OSS Helper

A collection of skills for AI coding agents (Claude, Bob, Gemini, OpenCode, Codex) that provides guidelines for contributing to open source projects. The skills auto-detect the current project via git remote get-url origin and load project-specific configuration from rule files.

For agents that support skills natively (Claude, Bob, Codex), the helper is installed as 6 focused, user-invocable skills — just describe what you want to do (or invoke /oss-* directly) and the agent follows the appropriate guidelines automatically. For other agents (Gemini, OpenCode), individual commands are generated from the same guidelines.

Getting Started

Any project can use the helper by adding an .oss-ai-helper-rules/ directory to its repository root with three rule files:

my-project/
├── .oss-ai-helper-rules/
│   ├── project-info.md          # Repository URLs, issue trackers, related repos
│   ├── project-standards.md     # Build tools, commands, code style restrictions
│   └── project-guidelines.md    # Branch naming, commit formats, PR policies
└── ...

Use the Add Project guideline to generate initial rule files for any project. Once committed, every contributor gets the right configuration automatically — no per-user installation of project-specific rules required.

Installation

Quick Install (All Agents)

curl -fsSL https://raw.githubusercontent.com/Open-Harness-Engineering/ai-agents-oss-helper/main/install.sh | bash

Selective Install

# Clone the repository
git clone https://github.com/Open-Harness-Engineering/ai-agents-oss-helper.git
cd ai-agents-oss-helper

# Install for specific agent
./install.sh claude    # Claude only
./install.sh bob       # Bob only
./install.sh gemini    # Gemini CLI only
./install.sh opencode  # OpenCode only
./install.sh codex     # Codex only
./install.sh           # All agents

How It Works

The helper provides guidelines that are project-agnostic. Project-specific configuration is stored in rule files with three files per project:

  • project-info.md - Repository URLs, issue trackers, SonarCloud keys, related repos
  • project-standards.md - Build tools, commands, code style restrictions
  • project-guidelines.md - Branch naming, commit formats, PR policies, task labels

Skill groups

The 37 guidelines are organized into 6 focused skills:

  • /oss-issues — Issue lifecycle: fix, analyze, create, list, find tasks, triage, cross-repo issues
  • /oss-review — PR lifecycle: review, address feedback, batch review, status, merge, backport
  • /oss-ci — CI/CD and code quality: fix CI errors, SonarCloud, GitHub alerts, quick fixes
  • /oss-security — Security: triage reports, drive a report end to end, analyze CVEs, draft advisories, sync the GitHub Advisory Database, scan code
  • /oss-project — Project setup: add projects, install/generate/update rules, workspaces
  • /oss-qe — Quality engineering: create and execute test plans, deep module audits

Each skill is user-invocable — you can invoke it directly via /oss-issues, /oss-review, etc. or the agent can auto-invoke it when your request matches. Individual /oss-* commands (e.g. /oss-fix-issue) are also available as thin wrappers that delegate to the appropriate skill.

Rule loading priority

Each skill initializes by loading project rules in this priority order:

  1. Project-local rules - .oss-ai-helper-rules/ directory in the repository root. Highest priority, versioned with the project.
  2. Installed fallback rules - A subdirectory under the agent’s local rules directory (for example ~/.claude/rules//) whose project-info.md declares a matching Remote pattern. Install these on demand with the Install Info guideline.
  3. Auto-discovery - If no rules exist anywhere, the agent auto-discovers the project’s configuration (build tool, conventions, etc.) and generates rule files in .oss-ai-helper-rules/ so they can be committed and shared.

Projects should adopt project-local rules so that configuration travels with the repository and stays in sync across all contributors and agents.

Where the rules come from

Project rules are not bundled with the installer. They live in a separate repository, Open-Harness-Engineering/ai-agents-oss-known-projects, and are installed on demand via the Install Info guideline. This keeps the helper focused on guidelines and lets projects that prefer not to host AI-agent metadata in their source tree have rules hosted centrally instead.

Capabilities

The OSS Helper provides guidelines for the following tasks. For agents with skill support (Claude, Bob, Codex), just describe what you want in natural language or use the /oss-* command directly. For other agents, use the corresponding command.

Capability Command Description
Fix an issue /oss-fix-issue Fix an issue from the project’s tracker (GitHub or Jira)
Review a PR /oss-review-pr Review a pull request against project rules and contribution standards
Find a task /oss-find-task Find an issue to contribute based on experience level
Create an issue /oss-create-issue Create a new issue in the project’s issue tracker
Quick fix /oss-quick-fix Apply a quick fix without a tracked issue (CI, docs, deps, etc.)
Analyze an issue /oss-analyze-issue Analyze an issue to understand the problem and investigate the codebase
Fix SonarCloud issues /oss-fix-sonarcloud Fix SonarCloud issues for a given rule
Fix GitHub alert /oss-fix-github-alert Fix a GitHub Code Scanning, Dependabot, or Secret Scanning alert
Add a project /oss-add-project Add a new project with the helper
Update knowledge /oss-update-knowledge Update a project’s rule files from a description or URL
Fix CI errors /oss-fix-ci-errors Download CI build reports, identify errors, and fix them
Fix backlog task /oss-fix-backlog-task Fix a task from a Backlog.md file (requires Backlog MCP server)
PR status /oss-pr-status Check CI checks, review state, and merge readiness of a PR
List PR status /oss-list-pr-status List all your open PRs with CI, review, and merge readiness summary
List PRs /oss-list-prs List all open PRs in the repo for browsing and review selection
List issues /oss-list-issues List issues assigned to you in the project’s tracker
Backport a PR /oss-backport-pr Cherry-pick a merged PR onto a maintenance/release branch
Address review /oss-address-review Address review feedback on a PR
Merge a PR /oss-merge-pr Merge a PR after verifying all requirements are met
Triage security report /oss-triage-security-report Triage an inbound security vulnerability report
Drive a security case /oss-security-case Drive one security report through triage, reporter reply, CVE, sanitized issue, fix, backports and CVE draft, stopping at every maintainer gate (a Claude Code workflow; other agents follow the same graph by hand)
Draft CVE advisory /oss-draft-cve Draft a project-specific CVE advisory page plus the CVE Record Format JSON
Analyze third-party CVE /oss-analyze-third-party-cve Analyze exposure to a CVE in a third-party dependency
Create security advisory /oss-create-security-advisory Privately report a security vulnerability via GitHub
Update GH Advisory Database /oss-update-gh-advisory-db Populate GitHub Advisory Database entries for the project’s published CVEs
Triage an issue /oss-triage-issue Triage a filed issue: reproduce, dedupe, classify, recommend disposition
Review batch of PRs /oss-review-prs Review a batch of open PRs you haven’t reviewed yet
Security scan /oss-security-scan Scan first-party code for security vulnerabilities
Generate project rules /oss-create-rules Generate project rule files for a new or existing repository
Install project rules /oss-install-info Install project rules from the known-projects repository
Initialize workspace /oss-workspace-init Initialize or rediscover a multi-repo workspace
Workspace status /oss-workspace-status Report status of all repos in a workspace
Create multi-repo issue /oss-create-multi-repo-issue Create and link issues across multiple repositories
Fix multi-repo issue /oss-fix-multi-repo-issue Fix an issue spanning multiple repositories
Create test plan /oss-qe-create-test-plan Create a test plan for a project feature or component
Execute test plan /oss-qe-verify Execute an existing test plan and track results
Audit a module family /oss-qe-module-audit Deep-audit a module family for bugs, antipatterns, inconsistencies, missing features, and ideas

Usage Examples

Fix an Issue

# With Claude, Bob, or Codex — just ask naturally:
fix issue 42
fix issue CAMEL-20410
fix https://github.com/wanaku-ai/wanaku/issues/42

# With Gemini or OpenCode — use the command:
/oss-fix-issue 42
/oss-fix-issue CAMEL-20410

Find a Task

# Natural language:
find me a task to contribute to

# Command:
/oss-find-task

Review a Pull Request

# Natural language:
review PR 42
review https://github.com/wanaku-ai/wanaku/pull/42

# Command:
/oss-review-pr 42

Quick Fix

# Natural language:
upgrade Quarkus BOM to 3.18.0
fix broken link in CONTRIBUTING.md

# Command:
/oss-quick-fix upgrade Quarkus BOM to 3.18.0

Backport a Merged PR

# Natural language:
backport PR 42 to the release/1.x branch

# Command:
/oss-backport-pr 42 branch=release/1.x

Drive a Security Report End to End

# Start a case from a report file; it stops at the verdict gate:
/oss-security-case xml-entities report=/path/to/report.txt

# Each later run records your decisions and approvals, and advances to the next gate:
/oss-security-case xml-entities verdict=VALID approve=issue
/oss-security-case xml-entities cve=CVE-2026-12345 approve=pr
/oss-security-case xml-entities done=reply approve=backports

# Once every fixed version is released, the next run drafts the advisory and the CVE record,
# kept in the case directory and archived only to your private draft repository, if you set one:
/oss-security-case xml-entities

Every run leaves a STATUS.md in the private case directory listing what is done, what waits on you, and the next command. Nothing is sent, reserved, merged, signed or published by the agent: those stay maintainer actions. The project’s project-security.md supplies the CNA, the advisory template and the supported lines; see skills/oss-security/security-case.md for the other settings.

Install Project Rules

# Natural language:
install project rules for camel-core

# Command:
/oss-install-info camel-core
/oss-install-info auto           # detect from git remote
/oss-install-info all            # install all known projects

Pointing at a different rules repository

The default source is Open-Harness-Engineering/ai-agents-oss-known-projects on the main branch. You can override either part with environment variables — useful for teams that maintain a private fork of the rules repo:

Variable Default Purpose
OSS_KNOWN_PROJECTS_REPO Open-Harness-Engineering/ai-agents-oss-known-projects GitHub org/repo to read rules from
OSS_KNOWN_PROJECTS_BRANCH main Branch within that repository

Agent-Specific Notes

Claude, Bob

Installed as 6 user-invocable skills at ~/.{agent}/skills/oss-{issues,review,ci,security,project,qe}/ plus individual /oss-* commands in ~/.{agent}/commands/. You can either describe what you want in natural language (the agent matches it to a skill automatically), invoke a skill group (e.g. /oss-issues), or invoke a specific command directly (e.g. /oss-fix-issue 42).

For Claude only, the installer also copies the workflows in workflows/ to ~/.claude/workflows/. These are scripts for Claude Code’s Workflow tool, which runs the subagents of a multi-step guideline in the background. /oss-security-case uses one. Claude Code reads the directory when a session starts, so start a new session after installing. Bob installs the same guideline and follows it by hand.

Gemini CLI

Individual TOML commands are generated from each guideline file and installed to ~/.gemini/commands/. Each command includes a preamble instructing Gemini to read project rules from ~/.gemini/rules//.

OpenCode

Individual markdown commands with frontmatter are generated and installed to ~/.config/opencode/commands/. Each command includes a preamble for project rule loading from ~/.config/opencode/rules//.

Codex

Installed as 6 skill directories at ~/.agents/skills/oss-{issues,review,ci,security,project,qe}/ with all guideline files as supporting files. Project rule files are installed to ~/.codex/oss-helper/rules/.

If Codex reports a missing ~/.codex/oss-helper/.oss-init.md, rerun ./install.sh codex from an updated checkout, then start a new Codex session. The installer removes stale standalone OSS Helper skills; current skills use their bundled init.md.

Project Structure

ai-agents-oss-helper/
├── install.sh                              # Installation script
├── README.md
├── workflows/                              # Claude Code workflows (installed for Claude only)
│   └── oss-security-case.js               # Security report lifecycle graph
├── tests/
│   └── workflows/
│       └── oss-security-case.test.mjs     # Routing test with mocked agents
└── skills/
    ├── _shared/
    │   └── init.md                         # Shared initialization logic
    ├── oss-issues/                          # Issue management skill
    │   ├── SKILL.md                        # Skill entry point
    │   ├── fix-issue.md                    # Fix an issue
    │   ├── analyze-issue.md               # Analyze an issue
    │   ├── create-issue.md                # Create a new issue
    │   ├── list-issues.md                 # List assigned issues
    │   ├── find-task.md                    # Find a task to contribute
    │   ├── fix-backlog-task.md            # Fix a backlog task
    │   ├── oss-triage-issue.md            # Triage a filed issue
    │   ├── oss-create-multi-repo-issue.md # Create cross-repo issue
    │   └── oss-fix-multi-repo-issue.md    # Fix cross-repo issue
    ├── oss-review/                          # PR management skill
    │   ├── SKILL.md
    │   ├── review-pr.md                    # Review a PR
    │   ├── address-review.md              # Address review feedback
    │   ├── oss-review-prs.md              # Review batch of PRs
    │   ├── pr-status.md                   # Check PR status
    │   ├── list-pr-status.md              # List your PR statuses
    │   ├── list-prs.md                    # List open PRs
    │   ├── merge-pr.md                    # Merge a PR
    │   └── backport-pr.md                 # Backport a merged PR
    ├── oss-ci/                              # CI/CD & quality skill
    │   ├── SKILL.md
    │   ├── fix-ci-errors.md               # Fix CI errors
    │   ├── fix-sonarcloud.md              # Fix SonarCloud issues
    │   ├── fix-github-alert.md            # Fix a GitHub alert
    │   └── quick-fix.md                    # Apply a quick fix
    ├── oss-security/                        # Security skill
    │   ├── SKILL.md
    │   ├── triage-security-report.md      # Triage security report
    │   ├── security-case.md               # Drive a security report end to end
    │   ├── analyze-third-party-cve.md     # Analyze third-party CVE
    │   ├── draft-cve.md                   # Draft CVE advisory
    │   ├── create-security-advisory.md    # Create security advisory
    │   ├── update-gh-advisory-db.md       # Sync GitHub Advisory Database
    │   └── oss-security-scan.md           # Scan codebase
    ├── oss-project/                         # Project setup skill
    │   ├── SKILL.md
    │   ├── add-project.md                 # Add a new project
    │   ├── install-info.md                # Install project rules
    │   ├── oss-create-rules.md            # Generate project rules
    │   ├── update-knowledge.md            # Update project rules
    │   ├── oss-workspace-init.md          # Initialize workspace
    │   └── oss-workspace-status.md        # Workspace status
    └── oss-qe/                              # Quality engineering skill
        ├── SKILL.md
        ├── oss-qe-create-test-plan.md     # Create test plan
        ├── oss-qe-verify.md               # Execute test plan
        └── oss-qe-module-audit.md         # Deep module-family audit

Project rule files are no longer bundled with this repository. They live in Open-Harness-Engineering/ai-agents-oss-known-projects and are installed on demand via the Install Info guideline.

Contributing

  1. Fork the repository
  2. Create a feature branch
  3. Add or modify guideline files in the appropriate skills/oss-*/ directory
  4. Update install.sh if adding new files
  5. For a Claude Code workflow, add the script to workflows/ (its meta.name must match the file name), list it in WORKFLOW_FILES in install.sh, give it a guideline that also works by hand for the other agents, and add a test under tests/workflows/
  6. Submit a pull request

License

Apache License 2.0


GitHub Repository

View this README on GitHub

Recommended Tools

Try a different keyword or remove a filter.

Install

npx skillfish add open-harness-engineering/ai-agents-oss-helper