Documentation: SKILL.md should emphasize fetching fresh data

[joseph-eldo]

Documentation: SKILL.md should emphasize fetching fresh data

Issue

When LLMs use the SlopeSniper skill in conversation, they may rely on cached data from earlier in the chat instead of running fresh commands. This leads to stale price quotes and inaccurate position values.

Example

User: "How's my Peyote position looking?"

What happened:

• LLM recalled price from 30 minutes ago: $0.00014 (~$8.81 position)
• Actual current price: $0.000094 (~$6.71 position)
• User saw wrong value and had to explicitly ask for a fresh check

Root Cause

Crypto markets move fast, but LLMs naturally prefer using existing conversation context over running new commands. The SKILL.md doesn't explicitly warn against this behavior.

Proposed Fix

Add prominent guidance in SKILL.md:

## Important: Always Fetch Fresh Data

**CRITICAL:** When the user asks about prices, positions, or market data:
- **ALWAYS run the command** - Don't rely on cached conversation data
- Crypto markets move fast - data older than 30 seconds is stale
- Run `slopesniper price MINT` or `slopesniper wallet` every time

**Example:**
User: "How's my Peyote position?"
❌ WRONG: Use price from 30 minutes ago in conversation history
✅ CORRECT: Run `slopesniper wallet` to get current value

Place this section prominently near the top, before command listings.

Additional Improvements

Consider adding timestamps to command output:

{
  "price_usd": 0.000094,
  "market_cap": 93779,
  "timestamp": "2026-01-27T23:37:35Z",
  "age_seconds": 0
}

This helps the LLM detect stale data and know when to refresh.

Impact

• User trust - Stale prices undermine confidence
• Trading decisions - Wrong data = wrong trades
• Debugging time - Users have to manually verify everything

Related

This is particularly important for:

• Price checks (slopesniper price)
• Wallet balance (slopesniper wallet)
• Market data (slopesniper scan)
• Position tracking (when we add PnL tracking per Feature Request: Built-in PnL Tracking and Portfolio Analytics #25)

Less critical for:

• Configuration (slopesniper config)
• Strategy settings (slopesniper strategy)
• Static data (mint addresses, transaction signatures)

Test Case

Setup: User buys token at $0.10
Wait: 30 minutes (price drops to $0.05)
User asks: "How's my position?"

Expected: LLM runs slopesniper wallet, reports current $0.05 price
Bug: LLM recalls old $0.10 price from conversation, reports wrong value

---

Priority: Medium-High (affects accuracy of all price-related queries)