Fetching from the wire…
Skills2026-07-30 · source-backed
Ten skills inlined at ~2,000 tokens each burns 20,000 tokens on knowledge the agent mostly doesn't need. Carry only names and one-line descriptions in the system prompt at roughly 100 tokens per skill, and expose a load_skill(name) tool that returns the full body wrapped in a tag when the model asks. Documented in a 72,734-star teardown alongside context compaction, subagent spawning, and worktree-isolated parallel execution as separately implementable units.
Each link below shares sources, entities, or timing with this story.
Shared entity: Split / Shared topic / Earlier coverage / Tension
Both cover Split; overlapping topics (agent, compaction, context, each, token); earlier Split coverage from 2026-07-22.
Same source / Shared topic
Cite the same source (Documented in a 72,734-star teardown); overlapping topics (agent, context, description, prompt, token).
Shared entity: Split / Same source domain / Shared topic / Earlier coverage
Both cover Split; reported by the same outlet (github.com); overlapping topics (agent, demand, token).
Same source domain / Shared topic / Tension
Reported by the same outlet (github.com); overlapping topics (agent, context, execution, prompt); pushes against this story (against).
Reported by the same outlet (github.com); overlapping topics (agent, prompt, token); pushes against this story (versus).
Reported by the same outlet (github.com); overlapping topics (agent, alongside, context); pushes against this story (but).
Reported by the same outlet (github.com); overlapping topics (agent, prompt, skill); pushes against this story (but).
Shared topic / Tension
Overlapping topics (agent, context, each, skill, token); pushes against this story (vs).