Headroom

Headroom is the transport layer: it compresses fresh tool output and new turns before they reach the model. It is not a replacement for source navigation, code review, or a model's actual context-meter.

When to use this skill

• Install or update Headroom's CLI, proxy, persistent deployment, wrapper, or MCP server.
• Verify whether an agent is actually routed through a healthy Headroom proxy.
• Configure the code-work policy: Graphify preflight before source mutation and Ponytail minimization only once an authoritative host context percentage reaches 60.
• Diagnose provider routing, proxy health, token savings, or a failed deployment.

When not to use this skill

• The task is source or symbol discovery without compression concerns → use graphify or codebase-search.
• The task needs a code-size reduction review without a context policy → use ponytail.
• The task is generic agent configuration ownership → use agent-configuration.