vincelele/ai-fomo-skills/ai-fomo

AI FOMO

Purpose

Turn AI information overload into personally aligned judgment. This skill should produce a decision, not just a summary.

Use When

Use this skill when the user:

• sends AI-related content or a link
• points to raw/inbox/
• asks whether something is worth reading
• asks whether a podcast or video is worth listening to
• asks for a summary plus recommendation
• asks for a podcast learning note that can replace listening to the full episode
• asks whether something should go into wiki
• asks for signals or a digest from retained material
• gives feedback on filtering or summarization behavior

Do Not Use When

Do not use this skill for:

• initializing a new workspace; use ai-fomo-init
• connecting source APIs or feed collectors; use ai-fomo-sources
• non-AI content with no relation to the user's alignment layer
• generic summarization without value judgment

Required Context

Before judging material, read:

1. self-context/index.md if it exists
2. relevant self-context/preferences/* files when needed
3. wiki/index.md when existing knowledge may already cover the topic

If the Personal Alignment Layer is missing or too vague, ask for minimal clarification or recommend running ai-fomo-init.

Core Workflow

1. Read the source content directly or read the provided raw snapshot.
2. Read the Personal Alignment Layer.
3. Use references/alignment-rubric.md for value judgment.
4. Reply in chat with a complete but concise summary.
5. Make a three-tier recommendation:

• write now: high quality, high relevance, durable judgment
• ask first: useful but uncertain
• skip: weak, generic, hype-heavy, or low relevance

1. If writing files, use references/filing-rules.md.
2. Always write or update wiki/sources before checking themes or dossiers.
3. Prefer updating existing themes or dossiers over creating new pages.
4. Only write signals or digests when the user asks or when the workflow explicitly calls for it.
5. Record feedback as long-term preference only when the user indicates it should persist.

Podcast Learning Note Mode

Use this mode when the source is a podcast episode, video transcript, or long-form interview and the user wants to learn without listening end to end.

Before writing the note, read references/podcast-learning-note.md. Use assets/podcast-learning-note.template.md as the scaffold when creating a Markdown file.

Default output target:

digests/podcasts/YYYY-MM-DD-short-slug.md

The note must answer three decisions:

1. What is this episode about, and is it worth continuing?
2. If continuing, what does the full episode cover in a readable learning structure?
3. If listening to the original, which segments are worth hearing and which can be skipped?

Keep podcast learning notes separate from durable source summaries. Use digests/podcasts/ for human-readable learning notes and wiki/sources/ for retained long-term knowledge.

Default Chat Output

For each source, include:

• what it is about
• main structure or claims
• durable mechanisms, constraints, or tradeoffs
• why it matters or does not matter for this user
• recommendation: write now, ask first, or skip
• proposed filing target when relevant

Judgment Rules

• Novelty is not importance.
• Hype is not signal.
• Long content is not automatically worth filing.
• Long audio is not automatically worth listening to end to end.
• Good filing candidates change product, system, market, evaluation, or workflow judgment.
• If a source mainly reinforces an existing theme, update that theme rather than creating a new one.
• Do not create a new theme for a one-off concept.
• For podcasts, distinguish "worth reading the learning note", "worth selected listening", and "worth full listening".

References

• Use references/alignment-rubric.md for judging value.
• Use references/filing-rules.md before writing workspace files.
• Use references/podcast-learning-note.md before writing podcast learning notes.
• Use references/examples.md for calibration or testing.