What problem does it solve? People cannot see their own repeated work: studies show self-reported computer use misses logged use by 32% on average. This Skill sweeps your Littlebird capture for recurring manual workflows, dedupes them against the skills you already own, and drafts a validated SKILL.md for the one candidate actually worth building. ## Core Features & Use Cases - Five-signature detection: Sweeps 90 and 180 days of capture for repeated setup work, manual data movement between apps, rebuilt documents, periodic tasks, and repeated requests like "same as last time", each with dated receipts. - Dedupe against installed skills: Classifies every candidate into already-covered, covered-but-not-firing, nearly-covered, or genuinely new buckets, often answering "you already have this, it needs better triggering" with your own quoted phrasings. - Honest build-or-skip ranking: Applies ten reasons not to automate, reports counts as lower bounds ("at least N"), and treats zero proposals as a complete result. - Use Case: Run it monthly via its self-installed routine, then open a session to review the ranked shortlist and receive a validated draft SKILL.md for the top candidate. ## Quick Start Ask the assistant to run skill-suggester and find repeated manual work in my Littlebird capture from the last 90 days.
