find-oss-alternative

name	find-oss-alternative
description	Finds the best open-source alternatives to a product, SaaS tool, feature, module, architecture, or skill before an agent builds from scratch. Use when a user asks for OSS alternatives, open-source competitors, replacement targets, architecture precedents, or a Use / Borrow / Avoid / Build New recommendation.

Find OSS Alternative

Find the best open-source alternative before building from scratch.

This skill uses public open-source projects and OSS alternative directories as evidence before important product, architecture, implementation, or skill-design decisions. The workflow finds credible alternatives, verifies the repos, studies real code, checks license and maintenance signals, credits builders, and turns the evidence into a practical implementation plan.

Default Workflow

1. Clarify intent

   • Write one sentence for the user's actual goal.
   • Identify product, workflow, architecture, implementation, protocol/model, analogy, and negative keywords.
   • Do this even when the user is still at idea stage.

2. Route precedent lanes

   • Skill precedents: search existing skills when the user may need a workflow or agent capability.
   • Open-source alternative directories: when the user names a product, SaaS, or category and wants an open-source replacement, use public alternative directories as candidate seed sources.
   • Repo precedents: search public GitHub repositories for open-source systems, modules, and implementation patterns.
   • Product/community precedents: search HN or the web for real-world naming, demand, pain, and product expectations.
   • Code evidence: use Sourcegraph or repo reads to verify that candidates really implement the relevant patterns.

3. Search with expanded queries

   • Do not rely on one phrase.
   • Search several query families: product category, user workflow, architecture, implementation anchors, analogies, and exclusions.
   • Prefer no-login tools first. See Toolchain.

4. Read the best candidates

   • Start with README, topics, license, stars, forks, update recency, and docs.
   • Treat directory results as leads only. Verify the GitHub repo, license, maintenance, and real code before using or borrowing from a candidate.
   • Deep-read only the most relevant projects.
   • Use Repomix, DeepWiki, local code search, or direct repo inspection when the candidate is likely to shape implementation.

5. Rank by practical fit

   • Fit to the idea matters more than stars.
   • Prefer projects with matching user workflow and implementation shape.
   • Penalize awesome lists, toy demos, abandoned repos, incompatible licenses, and projects that only match keywords but not intent.

6. Return a decision

   • Use: directly use or test this project/skill.
   • Borrow: learn product, architecture, workflow, or implementation patterns.
   • Avoid: do not copy this direction; explain why.
   • Build New: no precedent fits enough, or the user has a clear differentiating constraint.

Required Output

Always include:

• The intent as understood.
• The query families used or recommended.
• Ranked skill/repo/product precedents, separated by source type.
• Practical lessons for product and architecture.
• A Use / Borrow / Avoid / Build New decision.
• Credits and license notes for every repo that influenced the recommendation.
• Concrete constraints to carry into the next coding step.

Optional Outputs

Offer these when useful:

• HTML Dashboard for visual comparison and sharing.
• Precedent Reel for short-form demos. Do not treat video generation as a built-in dependency. If the environment has video skills or tools available, choose the best fit:
  • Use HyperFrames for fast HTML-first explainers, dashboard-to-video walkthroughs, and short social reels.
  • Use Remotion for React-first video projects, reusable templates, parameterized compositions, or a larger programmatic video system.
• Candidate JSON for another agent to continue the work from the same source of truth.

Use Output Modes for structure and Credits and Licenses for attribution rules.