Back to Skills
Optimization

Agent Slim

Vetted and production-tested agent skill.

Agent Slim

Category: Optimization

---

Audit, trim, and optimize an agent's always-resident context files without losing recall. Six-phase migrate-verify-slim sequence.

What It Does

Reduces resident token load by 75-90% while keeping all content searchable on demand. Moves archive content to searchable files, preserves backups at every step, enables prompt caching, and verifies nothing breaks before cutting.

The Process

1. **Backup everything** — Full workspace snapshot + git commit

2. **Migrate** — Move archive content to searchable files

3. **Verify** — Hard gate: representative recall queries must pass

4. **Slim** — Write lean resident files (~85% reduction)

5. **Cache** — Enable prompt caching on primary + sub-agent models

6. **Verify again** — 48h observation window for compaction failures

Safety Rules

  • Backup first — never cut without a snapshot
  • Original files NEVER deleted — renamed to `.bak`
  • Any verification failure → halt → report
  • All archive content stays searchable

Requirements

  • Workspace writable
  • `git` for version control
  • Provider that supports prompt caching (OpenRouter, Anthropic, OpenAI)

Get the full skills pack

Enter your email to download all 15 AgentVet skills as a pack. No spam, just occasional updates on new skills.

AgentVet provides independent security vetting. Skills are not affiliated with their original creators unless stated.