can-i-help

Contributor guidance for any project - matches developer skills to test gaps, stale docs, bugspots, and open issues.

Part of the agentsys ecosystem.

Point a contributor at a project and get specific, data-backed suggestions for where they can help. Collects project health signals automatically, asks about the developer's background, then matches skills to needs with concrete next steps.

Why this plugin

• Use this when onboarding a new contributor and need to match work to their skills
• Use this when looking for good-first-issue candidates backed by actual project data
• Use this when a project needs help but maintainers don't know where to direct people
• Use this when you want to contribute to an open-source project but don't know where to start

Installation

agentsys install can-i-help

Quick start

/can-i-help

Three phases run in sequence:

1. Collect (scripts/collect.js, no LLM) - gathers project metadata and contributor signals into one JSON file
2. Match (Sonnet agent) - asks about developer background, matches skills to project needs
3. Guide (interactive) - for each recommendation, reads relevant code and explains what to do

Contributor signals

Beyond base project data (manifest, structure, git), the collector gathers signals specific to contribution opportunities:

Signal	Source	What it reveals
Good-first areas	repo-intel can-i-help	Areas with clear patterns, low coupling, easy entry
Test gaps	repo-intel test-gaps	Hot source files without co-changing test files
Doc drift	repo-intel doc-drift	Documentation with low code coupling (likely stale)
Bugspots	repo-intel bugspots	Files with high bug-fix density
Stale doc references	repo-intel stale-docs	Doc lines that name symbols that no longer exist
Conventions	repo-intel conventions	Commit and naming style to match
Cleanup candidates	repo-intel slop-fixes	Orphan exports, commented-out code, passthrough wrappers, always-true conditions
Open issues	gh issue list	GitHub issues with labels and assignees

How matching works

The Sonnet agent tailors recommendations based on developer profile:

Developer says...	Recommended areas
"New to the stack"	Good-first areas with clear patterns and examples
"Experienced developer"	Hard problems in pain-point areas, architectural improvements
"I want to write tests"	Test gaps in frequently-changed files
"I want to fix bugs"	Bugspot files + relevant open issues
"I want to improve docs"	Stale documentation with concrete code examples

Each recommendation includes:

• What - specific file, area, or issue
• Why - data-backed reason (bug rate, test gap, staleness)
• How - reads relevant code and explains what needs doing
• First step - exact action to take

Usage

/can-i-help                        # Current repo, normal depth
/can-i-help /path/to/repo          # Specific repo
/can-i-help --depth=deep           # Include AST symbol data

Validated on 100 repos

The collector passes on 100 open-source repositories across 8 languages (JS/TS, Rust, Go, Python, C/C++, Java, Deno). Contributor signal queries validated on 8 repos with manual cross-checking against actual code: phantom directories filtered, source code prioritized over test/doc dirs, generated data excluded from recommendations.

Requirements

• Git repository with history
• agent-analyzer for contributor signals (optional; the collector builds the repo-intel map when it is missing)
• Node.js, to run the collector script
• GitHub CLI (gh) for open issues (optional)

Related plugins

• onboard - codebase orientation (what the project is and how it works)
• repo-intel - unified repository analysis (data source for all contributor signals)

License

MIT