Skill Locker
All skills
v1.3.03 loop iterations

Think Out Loud Router

You shouldn't have to know what your voice memo wants to be. The Think Out Loud Router auto-detects intent and routes it to one of six formats — blog, SOP, email, strategic memo, task list, or concept note — without you choosing.

What this skill does

People think in conversation, not in outlines. The best ideas show up in the car, in the shower, on a walk — when you're talking freely without editing as you go. But voice transcripts in their raw form are useless: repetition, tangents, half-finished thoughts, "um" and "you know" every other line, and circular reasoning where the same point gets made three times in different words. The gap between "I had a great idea while driving" and "here's a structured document I can act on" is where most ideas die. This skill closes that gap by routing automatically — you don't have to know whether your memo was a blog post or a memo or a task list.

The routing reads for signals. Decision signals ("I've decided", "the plan is") route to memo or task list. Teaching signals ("here's how to", "the way I think about this") route to blog or SOP. Communication signals ("I need to tell Sarah", "the message is") route to email. Exploration signals ("what if", "I'm wondering") route to concept note. Process signals ("every time we do this") route to SOP. Each of the six formats has its own extraction pattern — a blog post pulls out hook, core argument, supporting points, and implication; an SOP pulls out trigger, steps, decision points, and common pitfalls. Same input, different mining.

Repetition gets used, not discarded. When someone says the same thing three times in a voice memo, the second or third pass is usually the clearest expression — the skill picks the strongest version and uses that one. Tangents get classified: useful tangents get rehomed into a secondary output (action items, ideas worth saving), genuinely irrelevant ones get dropped. Incomplete thoughts get completed where the intent is clear and flagged [UNCLEAR] where it isn't.

It will not refuse to work. Even with a garbled transcript, the skill extracts what's salvageable, marks unclear sections, and produces a best-guess routing with a low-confidence flag. The user recorded a voice memo specifically because they didn't want to sit down and write — bouncing the work back as "this is too messy" defeats the point. One pass, no back-and-forth, and the speaker's voice gets preserved where it's strong. The output keeps your phrasing instead of corporate-ifying casual brilliance.

When this triggers

  • ·You recorded a voice memo while walking and don't know yet what it should become
  • ·Your brain dump contains a teachable idea, three to-dos, and an email you need to send
  • ·You want one skill that replaces five 'voice dump to X' variants
  • ·You're a verbal thinker and freeze when asked to outline before speaking
  • ·The transcript is messy enough that you'd accept partial output over 'I can't work with this'

Example

Trigger

User pastes 1,100 words of a voice memo transcript covering a pricing rethink, two tasks, and an idea for a podcast episode.

Output

Detected intent: Strategic Memo (high confidence) Secondary content: 2 action items, 1 concept note seed. Primary output — Strategic Memo Context: Pricing v1 is leaving money on the table at the top tier; revisiting structure before Q3 launch. Options analysed: 3-tier (current), 4-tier with enterprise, usage-based with floor. Recommendation: 4-tier with enterprise. Reasoning chain [reconstructed from the spoken arguments]. Next steps: pull cohort retention by tier, draft new pricing page. Secondary outputs Action items: · Pull cohort retention by tier — this week · Draft pricing page v2 in Figma — by next Fri Idea worth saving (concept note seed): · "Why every pricing page lies about its top tier" — possible podcast/post topic; needs 5 more minutes of thinking before it's a full concept. Filler stripped: ~28%. Tangents reclassified: 2.

Get this skill + 3 more

Get the full Voice-First pillar (4 skills) or the complete library.

Get the full stack — $299

What you get

  • 177-line SKILL.md, ready to drop into ~/.claude/skills/
  • Tested through 3 Karpathy-loop iterations (versions v1.0.0 → v1.3.0)
  • Triggers automatically when relevant — no command to remember
  • Lifetime updates as the skill is refined further

More from Voice-First

Browse the full library

297 skills across 31 categories. One purchase, lifetime updates.

See all bundles