Taxi

Handle everything for ground transportation, from price comparison to booking, tracking, disputes, and expense management.

Security Scan

The skill's requirements and instructions are internally consistent with a personal ground-transportation assistant; it is instruction-only, requests no credentials or installs, and its behavior matches its description.

Purpose & Capability

Name/description match the content: all guidance and workflows relate to ride comparison, booking assistance, tracking, disputes, and expense logging. No unrelated credentials, binaries, or install steps are requested.

Instruction Scope

SKILL.md instructs the agent to read/write a ~/taxi/ folder (memory.md, trips.md, accounts.md, promos.md) and to drive browser interactions with ride-booking sites and public promo sources. That is coherent with the stated purpose, but it grants the skill authority to store and read local personal data (addresses, account emails, trip history). The instructions explicitly say not to store full payment details, which mitigates some risk, but the agent will still handle sensitive trip and account info if enabled.

Install Mechanism

No install spec and no code files — instruction-only. This minimizes disk footprint and eliminates remote package/install risks.

Credentials

The skill declares no environment variables, no credentials, and no config paths beyond the user-local ~/taxi/ directory. The fields suggested to be stored (emails, phone, credits) are reasonable for its purpose but could be sensitive; the skill does not request broad or unrelated credentials.

Persistence & Privilege

Persistence is local to ~/taxi/ (user-owned directory). The skill is not always-enabled and does not request elevated or cross-skill privileges. Still, storing trip history and account notes on disk is persistent — users should be aware of privacy implications and filesystem permissions.

Assessment

This skill appears to do what it says: compare rides, help with bookings, track trips, and manage disputes. It asks for no external credentials or installs, but it does create and use files under ~/taxi/ that can contain personal data (saved addresses, emails, trip receipts, referral codes). Before enabling: (1) inspect ~/taxi/ contents and set strict file permissions (chmod 700 or equivalent); (2) avoid putting full payment credentials or unencrypted tokens in the memory files; (3) confirm you are comfortable with the agent performing browser interactions on ride websites; and (4) if you want stronger protection, keep sensitive account credentials out of the skill's local memory and handle final payment/verification steps yourself.

When to Use

User needs ANY help with taxis, rideshares, car services, or ground transportation. Agent handles research, comparison, booking assistance, price monitoring, expense tracking, and problem resolution.

Architecture

Memory lives in ~/taxi/. See memory-template.md for setup.

~/taxi/
├── memory.md          # Preferences, saved addresses, default apps
├── trips.md           # Trip history and expenses
├── accounts.md        # Which apps user has, preferences per app
└── promos.md          # Active promo codes, referral links

Quick Reference

Topic	File
Memory setup	memory-template.md
City/app coverage	cities.md
Fare estimation	fares.md
Booking workflows	booking.md
Problem resolution	problems.md

Core Rules

1. Know User Context First

Before any recommendation, check ~/taxi/memory.md for:

• Preferred apps and why
• Saved addresses (home, work, airport)
• Payment preferences (card, cash, business)
• Special needs (wheelchair, car seat, pet-friendly)
• Budget sensitivity

2. Proactive Comparison

Task	Agent Action
User needs ride	Compare 3+ options with current prices
Surge detected	Alert user, suggest waiting or alternatives
Airport transfer	Compare rideshare vs shuttle vs taxi vs car service
Group ride	Calculate if multiple rides or XL is cheaper
Regular route	Track prices over time, suggest best booking window

3. Maximize Savings

• Search for active promo codes before any booking
• Check if user has referral credits in any app
• Compare scheduled vs on-demand pricing
• Suggest pool/shared when time permits
• Monitor surge and alert when it drops
• Find airport flat rates vs metered options

4. Assist Booking Process

What agent CAN do:

• Open booking websites/apps via browser
• Pre-fill pickup and destination
• Select vehicle type based on user needs
• Apply promo codes
• Set scheduled pickup time
• Navigate to payment confirmation step

What requires user:

• Final payment confirmation
• Entering new payment methods
• Phone verification

5. Handle Special Scenarios

Scenario	Agent Handles
Airport pickup	Check terminal, flight tracking for delays, meet & greet options
Business travel	Ensure receipt capability, corporate profile, expense categories
Late night	Prioritize safety features, well-rated drivers, tracked rides
Foreign country	Find local apps, check cash/card acceptance, language tips
Large group	Calculate split options, find vans/minibuses, coordinate multiple cars
Accessibility	Filter for wheelchair-accessible, verify with company if needed
With pets	Find pet-friendly options, carrier requirements
With kids	Car seat availability, family-friendly services
VIP/Executive	Car services, black car, chauffeur options
Event pickup	Coordinate pickup points, surge avoidance strategies

6. Track Everything

Log in ~/taxi/trips.md:

• Date, route, app used
• Price paid, promo applied
• Category (personal, business, specific project)
• Notes (late driver, surge, issues)

Provide:

• Monthly expense summaries
• Per-category totals
• Comparison vs public transit
• Year-over-year trends

7. Resolve Problems

When user has issue:

1. Identify problem type (overcharge, route, driver, safety, lost item)
2. Find correct channel (in-app, email, phone, social)
3. Draft complaint with relevant details
4. Track resolution status
5. Escalate if no response (BBB, local regulator, social media)

See problems.md for dispute workflows by app.

8. Stay Current

• Check for new apps entering user's city
• Monitor promo code sources (RetailMeNot, Honey, app-specific)
• Track regulatory changes (some cities ban/restrict services)
• Know seasonal patterns (holiday surge, airport construction)

9. Multi-Modal Integration

Sometimes taxi isn't best option:

• Compare with public transit for cost/time
• Suggest train + last-mile rideshare
• Rental car for multi-day or multi-stop
• Hotel shuttle for airport
• Walking + rideshare vs door-to-door

10. Privacy & Security

• Never store full payment details
• Mask home address in logs if user prefers
• Share trip tracking with emergency contacts
• Verify driver details before user enters vehicle

Common Traps

• Booking without checking surge → Always show current multiplier
• Assuming one app works everywhere → Check local alternatives first
• Ignoring scheduled ride option → Often 10-20% cheaper
• Not logging business trips immediately → Harder to categorize later
• Using first promo code found → Some stack, some conflict, check terms
• Recommending cheapest without context → Safety/reliability matter for some trips
• Forgetting airport rules → Designated pickup areas, terminal-specific pricing
• Not tracking flight delays → User waits, ride leaves or charges wait time
• Ignoring user's past preferences → Check memory before suggesting