High-Quality Info Sources

---

name: high-quality-info-sources

description: Build, curate, score, and maintain high-quality information source lists for AI, technology, business, or any topic. Use when the user asks to create a skill for trusted sources, make a watchlist of people/sites/accounts to follow, filter noisy sources into a smaller high-signal set, turn a link dump into a reusable monitoring system, or design a repeatable workflow for tracking official accounts, researchers, critics, and market signals.

---

# High-Quality Info Sources

Build a small, high-signal information radar instead of a giant attention landfill.

Core workflow

1. Clarify the monitoring goal.

2. Group sources by role, not by popularity.

3. Prefer primary sources over commentary.

4. Keep the default list small.

5. Add a review rule so the list stays useful.

Clarify the monitoring goal

Start by identifying what the user actually wants to track:

If the user does not specify, assume they want a balanced monitoring set with:

Group by source role

Do not return a flat pile of links unless explicitly requested. Organize sources into roles such as:

Default ordering:

1. official / primary

2. builders / operators

3. explainers

4. critics / risk voices

5. aggregators

Quality filter

Prefer sources that satisfy most of these:

Penalize sources that are:

Output patterns

Choose one of these depending on the request.

1. Small radar list

Use for users who want the minimum viable watchlist.

Format:

Aim for 8-15 sources.

2. Extended source map

Use when the user wants broad coverage.

Format:

3. Monitoring system

Use when the user wants an operational workflow.

Include:

Maintenance rules

When building a reusable source system, include these rules:

AI-specific default lens

When the user asks for AI information sources and gives no stronger constraint, combine:

Read `references/ai-sources.md` for a starter set and selection logic.

Tone and judgment

Be opinionated. A source list is a filter, not a census.

Prefer:

Avoid pretending all sources are equally good.