Find a paper, verify its DOI, classify its title
by SuperpowersOfficial1 install0 runs
Searches scholarly metadata, resolves the top result's DOI to its source record, then classifies the title with text moderation. Free endpoints; moderation is not fact checking.
Give it to your agent
Paste this into Codex, Claude Code or any agent with a terminal. It reads the recipe, sets up Superpowers if needed and runs it on your balance.
Use the Superpowers recipe at https://superpowers.tools/recipes/sourced-research-check/skill.md
Or run the script
curl -s https://superpowers.tools/recipes/sourced-research-check/run.mjs -o sourced-research-check.mjs node sourced-research-check.mjs
Reads your key from SUPERPOWERS_API_KEY or ~/.config/superpowers/key and prints the results link.
The recipe
Exactly what your agent reads.
--- name: sourced-research-check description: Find a scholarly paper on a topic with Superpowers, confirm its DOI resolves to the same record, and run the title through text moderation. Use when the user needs a citable source for a topic and wants the DOI checked. Free. --- # Find a paper, verify its DOI, classify its title A Superpowers recipe using three free tools. Moderation only classifies the title for content review; it is not fact checking. Needs a Superpowers API key. If you don't have one yet: https://superpowers.tools/skill.md?src=recipe%3Asourced-research-check ## Run the script curl -s "https://superpowers.tools/recipes/sourced-research-check/run.mjs" -o sp-research-check.mjs node sp-research-check.mjs --query "graph neural networks" ## Step by step 1. POST https://superpowers.tools/v1/run {"tool": "crossref.search-works", "input": {"query": <topic>, "limit": 2}} and take results[0].doi. 2. POST https://superpowers.tools/v1/run {"tool": "crossref.lookup-doi", "input": {"doi": <doi>}} and check the returned doi matches. 3. POST https://superpowers.tools/v1/run {"tool": "openai.moderate-text", "input": {"text": <title>}}. Report the title, the https://doi.org/ link and whether the DOI matched. Try the next search result if the user wants a different paper.