sk Skill · JuliusBrussee
cavecrew
When to delegate to cavecrew-investigator (locate code), cavecrew-builder (1-2 file edit) or cavecrew-reviewer (diff review) instead of working inline or using Explore. Their output is compressed, so main context lasts longer.
Open on skills.sh ↗read 2026-09-17
- installs 8w
- 100,275
- 30-day movement
- starts with the next reading
- Related entries
- 3
- Connections
- 0
Go
A SKILL.md that teaches an AI coding agent when to delegate work to cavecrew subagents (investigator, builder, reviewer) instead of working inline or using Explore. It defines output contracts, chaining patterns, and pitfalls, with subagent output compressed to preserve main-context tokens.
Reach for it when subagent tool results are bloating your main context and you want a rule of thumb for compressed delegation.
Use it to
- Decide between cavecrew and vanilla Explore
- Chain locate-fix-verify across subagents
- Learn what each cavecrew agent returns
- Avoid token-wasting delegation mistakes
For AI coding agent users managing context budgets
- Host repository
- JuliusBrussee/caveman
- Installs, lifetime
- 307k
- Installs, 8 weeks
- 100k
- Host stars
- 106k
- Host language
- Go
topicsagent-skillssubagentscontext-managementdelegationcode-review