adityahimaone/hermes-agent-rtk-caveman/context-optimization

Context Optimization Workflow

Use this skill when Adit wants to reduce token usage, optimize context, or check RTK/agent efficiency.

When to Use

• User asks about token savings, RTK coverage, or context management
• Periodic maintenance (monthly recommended)
• After noticing high session count or bloated database

Steps

1. Check Current State

rtk gain                          # RTK token savings
hermes insights                   # Full usage analytics (last 30 days)
hermes sessions stats             # Session store size

2. Prune Old Sessions

# Cron sessions older than 3 days (auto-generated, safe to delete)
hermes sessions prune --older-than 3 --source cron --yes
 
# Telegram sessions older than 14 days (if needed)
hermes sessions prune --older-than 14 --source telegram --yes

3. Verify RTK Enforcement

Check that CLAUDE.md contains:

• Golden Rule: always prefix with rtk
• MANDATORY section for execute_code terminal() calls
• Exception list (interactive, no-output, full-output)

If missing, patch CLAUDE.md with execute_code RTK rules.

4. Set Up Auto-Prune Cron (if not exists)

# Create cron: daily at 3 AM WIB
hermes sessions prune --older-than 3 --source cron --yes

Schedule: 0 3 * * * (daily 3 AM) Deliver: origin

5. Optimize Memory

• Check memory usage (% of 2,200 char limit)
• Consolidate redundant entries
• Update RTK mandate entry if changed

6. Report Summary

Present combined analysis:

• Sessions: before → after pruning
• RTK coverage: commands filtered / total terminal calls
• Memory: usage %
• Auto-prune: active/paused

Key Metrics to Track

Common Issues

• Low RTK coverage: Patch CLAUDE.md with stricter execute_code rules
• Memory at 100%: Consolidate or remove entries using memory tool replace action
• Session bloat: Run prune for cron source first (safest to delete)
• RTK name collision: Check which rtk — should be Rust Token Kit, not Rust Type Kit