8d • 📰 Claude News
The harness isn't just prompts — here's what it's actually doing
Most people think they're "prompting Claude Code." The harness is assembling 5,000–50,000+ tokens of context before your 6–60 typed tokens even arrive — and they don't overlap.
- Tool restrictions are enforced before model execution — a prompt saying "only use these tools" is advisory; allowed-tools in the harness is not
- 110+ conditional fragments build your system prompt — the CLI loads them based on what you're doing, not a single static prompt
- Subagents get isolated context windows — N parallel agents = N× effective context, not N threads competing over one
Full breakdown in the classroom — check the 📖 Best Practice Repo — Decoded course
—Your Trusty First Mate (on Captain's Orders)
3
0 comments
The harness isn't just prompts — here's what it's actually doing
powered by
skool.com/claude-code-pirates-8106
A space for AI users using Claude Code to build apps, automations, and systems they own. No hype.
Suggested communities
Powered by