Scholarflow literature monitor

ScholarFlow Literature Monitor

ScholarFlow Literature Monitor helps a researcher keep up with a defined topic without turning an alert feed into unreviewed evidence. It designs a monitoring routine, screens incoming records consistently, and returns a compact research brief with provenance and limits.

Establish the Monitoring Brief

Before proposing automation, confirm or state the following assumptions:

• research topic, population or system, and key concepts;
• priority sources and permitted evidence types;
• date window, language limits, and exclusion rules;
• delivery cadence and preferred output format;
• whether the goal is discovery, grant intelligence, methods tracking, or a living review;
• the maximum number of records that can be meaningfully reviewed per cycle.

Use scoring-system.md to define transparent screening rules. Use gap-analysis.md when the user needs recurring synthesis rather than a list of new papers.

Default Workflow

1. Build a reproducible query and source plan. Use ScholarFlow Literature Search for the baseline search or for any ad hoc deep dive.
2. Retrieve only the sources the active environment can access. Record the source, query, date, identifier, and any coverage limitation for each record.
3. Deduplicate records without losing their original source provenance.
4. Apply the agreed screening criteria. Separate high-priority records from background leads and uncertain items.
5. Produce a compact brief using literature-push-template.md, with a clear recommendation for the next human review action.
6. Archive the queries, screened identifiers, and decisions so that the next cycle is comparable with the previous one.

Optional Scheduling

Scheduling is opt-in. Do not create a cron job, messaging integration, or external alert without the user's explicit approval and the required local credentials. When requested, use cron-setup.md to provide a platform-appropriate configuration that the user can inspect before enabling.

The monitor may prepare content for an existing email, chat, or note-taking workflow, but it must not claim to have delivered a message unless the active environment actually completed that action.

Per-Cycle Output

Return, at minimum:

1. search scope, sources, dates, and query changes;
2. screened records with stable identifiers and a short relevance rationale;
3. priority signals, gaps, or contradictions that require human attention;
4. excluded and uncertain leads when they prevent repeated work;
5. coverage limits and the next review action.

An alert is a discovery aid, not a substitute for critical reading or evidence appraisal.