agentleFS
Martin-HausleitnerMartin-Hausleitner/martins-awesome-skills/skills/research/comparison-deep-research/SKILL.md

comparison-deep-research

Skill3 starsChanged 6 months agoMITView on GitHub

Use when the user asks to compare products, tools, GitHub repos, packages, frameworks, guides, platforms, plugins, or alternatives with Deep Research. Produces a strict research prompt and output format with at least 10 candidates, ideally 50, clickable GitHub and website links, five use-case-specific scoring categories, a 100-point total score, ranking, diagrams, and recommendation tables. Trigger on "vergleichen", "compare products", "best packages", "best tools", "GitHub links", "score 100", "rank platforms", "beste Programme", "Plugins vergleichen", or any broad comparison request that should become a reusable Deep Research scorecard.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 3 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this skill before starting Gemini/ChatGPT/Oracle Deep Research when the user wants products, repos, packages, tools, plugins, guides, …

    line 15
  2. Distinctive:

    Before submitting, check the prompt budget:

    line 44
  3. Very specific:

    Use --market-depth exhaustive when the user wants the market "ausgereizt" or as many candidates as possible.

    line 59
  4. Distinctive:

    Use exactly five top-level categories.

    line 97
  5. Very specific:

    Do not invent links, stars, licenses, or dates.

    line 101
  6. Very specific:

    Use ai-research-browser or ai-research-browser-cli for the browser run.

    line 107
  7. Very specific:

    Use docs/DEEP_RESEARCH_PROMPT_BUDGETS.md and the prompt-budget script before submitting any paid Deep Research run.

    line 108
  8. Very specific:

    Use --allow-paid-quota-use only when the user explicitly asked for Deep Research or a paid mode.

    line 109
  9. Very specific:

    Use ai-research-output-publisher after the report finishes to create a polished message, duration metrics, PSPin/deep-research links, and …

    line 111
  10. Very specific:

    Use tool-comparison-heatmap after the research if the user wants an interactive visual matrix from the final data.

    line 112
  11. Distinctive:

    Use-case categories might be:

    line 122

≈ “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

Comparison Deep Research

Use this skill before starting Gemini/ChatGPT/Oracle Deep Research when the user wants products, repos, packages, tools, plugins, guides, platforms, or alternatives compared.

The skill's job is to convert a vague comparison request into a strict Deep Research prompt that always asks for:

  • at least 10 candidates and ideally 50 when the domain is broad;
  • an exhaustive market mode when the user asks to "reize den Markt aus", "as many as possible", "vollständig", "alle relevanten", or similar;
  • verified GitHub repo links and official website/docs links;
  • a comparison table where names are clickable and link columns are explicit;
  • five major categories tailored to the use case;
  • a 100-point weighted score for each candidate;
  • category winners, overall winner, runner-ups, and "best for X" recommendations;
  • at least one diagram or visual summary suitable for later rendering.

Quick Use

Generate a prompt:

python3 skills/research/comparison-deep-research/scripts/comparison_deep_research_prompt.py \
  --topic "Discord voice agents and private assistants" \
  --use-case "private OpenClaw/Hermes-style Discord assistant that can speak live, run tasks, and post research results" \
  --candidate-count 50 \
  --market-depth exhaustive \
  --context-file /absolute/path/context-notes.md \
  --context-file /absolute/path/known-candidates.csv

Then pass the generated prompt into ai-research-browser or Oracle-assisted Deep Research.

Before submitting, check the prompt budget:

python3 skills/research/comparison-deep-research/scripts/deep_research_prompt_budget.py \
  --prompt "$(python3 skills/research/comparison-deep-research/scripts/comparison_deep_research_prompt.py --topic '...' --use-case '...' --candidate-count 50)" \
  --json

Ideal prompts stay at or below 6,000 characters. Default automated UI submits should stay at or below 12,000 characters. If the prompt is longer, compress it or move source material into files/links. Above 24,000 characters, do not paste it into Deep Research.

Market Exhaustive Mode

Use --market-depth exhaustive when the user wants the market "ausgereizt" or as many candidates as possible. In this mode, the prompt asks Deep Research to:

  • discover as many verifiable candidates as reasonable;
  • include long-tail and adjacent-category options;
  • build a longlist first, then a ranked shortlist;
  • state discovery queries and excluded-but-notable candidates;
  • still score the shortlist rigorously instead of producing an unreviewable dump.

Large context should be passed as files with --context-file and as workflow-run --attachment ... later. Do not paste full notes, exports, or large candidate lists into the prompt body if they push the prompt over the budget.

Required Deep Research Output

Tell the model to produce this structure:

  1. Executive Verdict: winner, top 3, and when not to use the winner.
  2. Candidate Discovery: explain search strategy and include excluded notable candidates.
  3. Scorecard Table:
    • Rank
    • Candidate as a Markdown link to the official website/docs
    • GitHub as a Markdown link when available, or N/A with reason
    • Type
    • License / Pricing
    • Last Activity
    • five tailored category scores
    • Total /100
    • Best For
    • Key Caveat
  4. Five Category Definitions: each category must have subcriteria and a fixed max score. The five categories must sum to 100.
  5. Mermaid Diagram: quadrant, flow, or decision tree that helps choose between candidates.
  6. Detailed Notes Per Candidate: concise evidence, links, and score rationale.
  7. Final Recommendation: MVP choice, production choice, budget choice, self-hosted choice, safest choice.
  8. Next Steps: concrete implementation or adoption plan.

Scoring Rules

  • Use exactly five top-level categories.
  • Assign category weights that sum to 100.
  • Score every candidate in every category.
  • Total score must be the weighted sum, shown as /100.
  • Do not invent links, stars, licenses, or dates. If a source cannot be verified, mark it Unverified and reduce confidence.
  • For GitHub projects, verify repo URL, license, last meaningful activity, release/commit freshness, ecosystem activity, and maintenance signals.
  • For non-GitHub products, verify official docs, pricing, security/compliance notes, and public roadmap or changelog.

Integration Guidance

  • Use ai-research-browser or ai-research-browser-cli for the browser run.
  • Use docs/DEEP_RESEARCH_PROMPT_BUDGETS.md and the prompt-budget script before submitting any paid Deep Research run.
  • Use --allow-paid-quota-use only when the user explicitly asked for Deep Research or a paid mode.
  • Attach large context files with --attachment /absolute/path/file when the target provider exposes a file input; keep the initial prompt as a compact brief.
  • Use ai-research-output-publisher after the report finishes to create a polished message, duration metrics, PSPin/deep-research links, and optional Notion payload.
  • Use tool-comparison-heatmap after the research if the user wants an interactive visual matrix from the final data.

Example Request Mapping

User: "Vergleiche die besten Obsidian AI Plugins."

Deep Research topic: "Best Obsidian AI plugins for rebuilding Notion-like team knowledge workflows."

Candidate count: 50 if available, minimum 10.

Use-case categories might be:

  • Notion-like knowledge management fit: 25
  • AI/RAG capability: 20
  • Team collaboration and sync: 20
  • Data ownership and privacy: 20
  • Maintenance, ecosystem, and ease of setup: 15

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.