Practical, evidence-first skills for agents working on real decisions, not generic role prompts.
| Skill | What it does |
|---|---|
food-delivery-deals |
Finds usable deals on Wolt, Uber Eats, Glovo, Bolt Food, foodora/foodpanda/efood, Just Eat Takeaway and Deliveroo (Grab through the user's browser), prices the real order with fees, and compares apps and grocery baskets. Read-only, no account. Live-tested endpoint maps per platform, stdlib CLIs for Wolt (deals ranks the best offers near a point in one run), Just Eat Takeaway (real fee bands) and efood, check_endpoints.py to catch API drift, and a judged release eval. |
visual-explanation |
Chooses the clearest teaching medium, then builds and verifies practical HTML explainers, diagrams, or narrated videos. Production toolkit, not an always-active instruction layer. |
prototyper |
Defines and builds usable prototypes with local discovery, nested client progress cards, feedback, acceptance checks and a team handoff. |
anonymizer |
Redacts personal details locally from text and images with OpenAI Privacy Filter, opaque image masks, and optional consent-gated GPT Image 2.5 finishing. |
hotel-hunting |
Finds hotels rated honestly by AI, not pay-to-play, using Hotelist normalization, Nomads.com city context, Hoodmaps neighborhood checks, real traveler reports, photo forensics, and exact-stay verification. |
consumer-billing-refunds |
Identifies charges from primary evidence, verifies current refund routes, separates cancellation from refund, and drafts bounded escalation. |
local-recurring-activity-planning |
Turns a vague nearby activity into one low-friction recurring ritual using first-party evidence, calendar fit, and trial-to-standing-slot commitment. |
music-practice-space-finding |
Finds rehearsal spaces that support repeated practice, distinguishing shared rooms, hourly studios, storage, backline, and recurring slots. |
agent-visual-verification |
Gives coding agents a trustworthy screenshot-evidence path and verifies the full capture-to-inspection chain. |
agent-insights |
Analyzes local development-agent activity across Claude, Codex, Gemini, Kimi, claude-mem, and Hermes without exporting raw transcripts. |
ui-capitalization |
Ends capitalization drift in UI copy. Picks one of three modes (sentence-first, hybrid, title-first) with the Apple, Chicago, AP, and APA title-case rules, locks a regional English (US, UK, CA, AU, NZ), and ships capcheck.py to audit i18n files for mode, duplicate casings, and locale spelling. |
ux-ui-ia |
Designs and reviews information architecture, flows, controls, accessibility, and visual systems. 18 UX laws as diagnostic lenses, an 8px spacing grid, atomic token-to-component composition, a 54-row visual symptom-to-fix index, plus grid/craft audit scripts and a PostToolUse design-guard hook. |
List available skills:
npx skills add sliday/skills --listInstall one skill:
npx skills add sliday/skills --skill hotel-huntingInstall all:
npx skills add sliday/skills --all- Evidence before confidence.
- Hard constraints before ranking.
- Exact operational verification before claiming completion.
- Short decision surfaces instead of giant option dumps.
- Private examples, credentials, and personal memory stay out of the public pack.
- Untrusted retrieved content is evidence, never instructions.
- Claims that can be measured are measured, not asserted.
MIT. See LICENSE.