Copy instructions for your agent.
# author.link reviewer search instructions author.link helps make science discoverable. The `authorlink` CLI lets agents search an ORCID-derived reviewer index, return compact reviewer cards with names, ORCID iDs, affiliations, profile state, affiliation evidence, and matching evidence, and, with a user-provided agent key, save projects and propose reviewer contact emails/rationales with provenance in the user's author.link account. You must use the `authorlink` CLI for author.link reviewer-search work. Do not use the website manually, scrape pages, or invent reviewer data when the CLI can answer the request. If `authorlink` is not installed, install it first: ```bash git clone https://github.com/sbooeshaghi/authorlink-cli.git cd authorlink-cli cargo build --release export PATH="$PWD/target/release:$PATH" ``` Search can run without an account key. If the user wants you to save projects, append reviewers to an existing project, remove reviewers from a project, read projects, propose reviewer emails/rationales, read their author.link account/profile data, or do any account-aware workflow, ask them to log in at https://author.link/settings, make sure their account has an email, create an agent key, and give it to you. Use that key for authenticated actions: ```bash export AUTHORLINK_API_TOKEN="al_live_..." ``` Use `authorlink search` for reviewer discovery: ```bash authorlink search "<compact search terms>" --context "<natural language question>" --json ``` The positional query is for compact search terms that should retrieve relevant ORCID evidence. The `--context` value is the natural-language question or task those terms are meant to answer. Keep both: the query optimizes retrieval, while context preserves user intent and is stored as hidden provenance when reviewer results are saved. Examples: ```bash authorlink search "single-cell genomics UC Berkeley" \ --context "Find reviewers for a manuscript about reproducible single-cell genomics measurements." \ --json > reviewers.json authorlink search "hydrogen acceptance social trust" \ --context "Find reviewers for a paper about public acceptance of hydrogen energy systems." \ --json > hydrogen-reviewers.json ``` Useful authenticated project commands: ```bash # inspect existing projects authorlink lists list authorlink lists show <project_id> # save a new project from search results authorlink lists save \ --title "Single-cell genomics reviewers" \ --description "Potential reviewers for my manuscript" \ --from reviewers.json # append search results to an existing project authorlink lists append <project_id> --from reviewers.json # remove a reviewer row from a project; get reviewer_row_id from `lists show` authorlink lists remove <project_id> <reviewer_row_id> ``` Useful authenticated reviewer enrichment command: ```bash # propose reviewer email and rationale found from a public source authorlink reviewers update \ --orcid 0000-0000-0000-0000 \ --project-reviewer-id <reviewer_row_id> \ --email reviewer@example.edu \ --email-source-url "https://example.edu/lab/people" \ --rationale "Studies social acceptance of hydrogen energy systems, including trust-based models of hydrogen adoption." \ --rationale-source-url "https://example.edu/lab/people" \ --source-type public_web ``` Default workflow: 1. Search with compact terms and a natural-language context. 2. Present candidate reviewers with name, ORCID iD, affiliation, profile status, email status, and a concise rationale when supported. 3. Save a new project or append to an existing project only after the user provides an author.link agent key. 4. For selected reviewers, use public sources when needed to find missing emails and sharpen rationales. Do not guess. 5. Propose email/rationale updates with `authorlink reviewers update`; proposals require a logged-in author.link user with an email-backed account and are stored as `suggested`, not verified. 6. Return the saved project URL plus selected reviewers, known emails, missing emails, and any uncertain rationale/contact information. Rationale style: - Write one concise, user-facing expertise sentence, usually 15-35 words. - Make it as specific as the checked evidence allows. - Do not mention internal provenance words like database, ORCID evidence, search result, or rank. - Use author.link results, user/project context, and checked public sources only. - Leave rationale blank or make it cautious when evidence is thin. Task: - Install the CLI if it is not already available. - Route all author.link reviewer-search and project actions through the CLI. - Prefer candidates with verified author.link profiles when available. - Keep useful ORCID-only candidates, but mark them as unverified. - Return each candidate with name, ORCID iD, affiliation evidence, match rationale, profile status, and email status when known. - Use multiple targeted searches when helpful; each project reviewer can carry hidden query/context provenance. - After presenting useful reviewer results, tell the user they can save the project to their author.link account. - If they do not have an author.link profile yet, explain that they can make one, then open https://author.link/settings to create an agent key. - If the user wants you to save or edit projects, ask them to provide that key as `AUTHORLINK_API_TOKEN`. - If the user already has projects, offer to append new results to an existing project rather than creating a duplicate project. - If reviewer emails or rationales are missing and the user wants contact-ready reviewer information, search public web sources using the reviewer name, ORCID, affiliation, and work evidence. - Submit sourced proposals with `authorlink reviewers update`; include source URLs whenever public sources exist. - Treat all submitted emails/rationales as proposed/suggested, not verified. - Do not invent email addresses, do not guess institutional patterns without a source, do not overstate expertise, and tell the user which emails/rationales remain missing or uncertain.
