agentleFS
rotinoM0rotinoM0/ai-opensource-finder/.claude/skills/project-recommender/SKILL.md

project-recommender

Skill0 starsChanged 3 months agoMITView on GitHub

Recommends open-source projects based on user profile and search history

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 16 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈16

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If GitHub username is set, analyze existing repos and contributions to avoid recommending known projects

    line 17
  2. Distinctive:

    Avoid recently recommended repos

    line 21
  3. Distinctive:

    Search configured sources (GitHub, Awesome lists)

    line 22
  4. Distinctive:

    When the user has a GitHub username in their profile:

    line 31
  5. Distinctive:

    Suggest complementary projects: if user has React repos, recommend projects in the React ecosystem they don't yet know

    line 36
  6. Distinctive:

    If user liked 3 projects with minimal APIs → boost "lightweight" and "minimal" in future scores

    line 43
  7. Distinctive:

    If user disliked projects with complex build setups → penalize "monorepo" or "ejected" patterns

    line 44
  8. Distinctive:

    If user consistently picks TypeScript over JavaScript → boost TypeScript variants

    line 45
  9. Distinctive:

    If user starts searching for a language they previously said they don't know → note the interest growth

    line 48
  10. Distinctive:

    If search queries shift from "frontend" to "full-stack" → update inferred career direction

    line 49
  11. Distinctive:

    If user explores a new ecosystem repeatedly → add it to inferred secondary skills

    line 50
  12. Very specific:

    Never recommend a repo the user already saw and rejected (check user_feedback.disliked)

    line 53
  13. Distinctive:

    Don't re-recommend the same repo within 7 searches (unless user asks)

    line 54
  14. Distinctive:

    If user rejected a category (e.g., "no WordPress plugins"), filter entire category

    line 55
  15. Distinctive:

    If past "React" queries returned too many boilerplates → add "production" or "used by" filter

    line 58
  16. Distinctive:

    If past "Python" queries returned ML-only results → diversify with "web", "cli", "automation"

    line 59
  17. Distinctive:

    If user liked projects with good first issues → prioritize repos with that label

    line 60

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Project Recommender Skill

Recommends open-source projects compatible with the user's profile, informed by past search history.

When to use

  • User asks to find projects to contribute to
  • User wants project recommendations based on their skills
  • User asks "what should I work on?"

Workflow

  1. Load user profile from CLAUDE.md and 01-user-profile.md
  2. If GitHub username is set, analyze existing repos and contributions to avoid recommending known projects
  3. Load search history from project_scraper/search_history.json:
    • Extract learned preferences (liked/disliked patterns)
    • Identify skill evolution (new interests emerging)
    • Avoid recently recommended repos
  4. Search configured sources (GitHub, Awesome lists)
  5. Score each project against evaluation criteria
  6. Apply history-based adjustments to scores
  7. Present ranked recommendations with fit analysis
  8. For user-selected projects, identify contribution opportunities
  9. Log the search and update history with user feedback

GitHub-aware recommendations

When the user has a GitHub username in their profile:

  • Exclude owned repos from recommendations (they already know their own projects)
  • Exclude repos they contribute to (already engaged)
  • Boost languages found in their repos but not in their stated skills (implicit expertise)
  • Cross-reference starred repos to identify ecosystem preferences
  • Suggest complementary projects: if user has React repos, recommend projects in the React ecosystem they don't yet know

History-aware recommendations

The search history acts as the agent's long-term memory:

Preference learning

  • If user liked 3 projects with minimal APIs → boost "lightweight" and "minimal" in future scores
  • If user disliked projects with complex build setups → penalize "monorepo" or "ejected" patterns
  • If user consistently picks TypeScript over JavaScript → boost TypeScript variants

Skill evolution

  • If user starts searching for a language they previously said they don't know → note the interest growth
  • If search queries shift from "frontend" to "full-stack" → update inferred career direction
  • If user explores a new ecosystem repeatedly → add it to inferred secondary skills

Avoidance patterns

  • Never recommend a repo the user already saw and rejected (check user_feedback.disliked)
  • Don't re-recommend the same repo within 7 searches (unless user asks)
  • If user rejected a category (e.g., "no WordPress plugins"), filter entire category

Query refinement

  • If past "React" queries returned too many boilerplates → add "production" or "used by" filter
  • If past "Python" queries returned ML-only results → diversify with "web", "cli", "automation"
  • If user liked projects with good first issues → prioritize repos with that label

Scoring criteria

  • Skills match (0-100): overlap between project stack and user skills (boosted by GitHub repo evidence and history patterns)
  • Activity level (0-100): recency of commits, issue response time
  • Community health (0-100): README quality, CONTRIBUTING.md, issue labels
  • Learning potential (0-100): exposure to new technologies, documentation quality
  • Contribution opportunity (0-100): open good first issue, unassigned issues

Score adjustments from history

SignalAdjustment
User liked similar stack+10 skills match
User disliked similar stack-15 skills match
User liked project size/complexity+5 activity/learning
User rejected project size/complexity-10 activity/learning
User explored this topic before+5 contribution opportunity
User explicitly rejected this category-50 overall (skip)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.