{"id":203,"date":"2026-02-21T06:47:32","date_gmt":"2026-02-21T14:47:32","guid":{"rendered":"https:\/\/chris.tsehome.com\/?p=203"},"modified":"2026-02-22T08:21:31","modified_gmt":"2026-02-22T16:21:31","slug":"the-million-token-anchor-mastering-sub-agent-efficiency","status":"publish","type":"post","link":"https:\/\/chris.tsehome.com\/?p=203","title":{"rendered":"The Million-Token Anchor: Mastering Sub-Agent Efficiency"},"content":{"rendered":"\r\n<h1 class=\"wp-block-heading\">The Million-Token Anchor: Mastering Sub-Agent Efficiency<\/h1>\r\n\r\n\r\n\r\n<p>In the world of AI automation, there is a hidden cost to curiosity. This week, while optimizing my daily news digest, I hit a wall: a &#8220;Million-Token Anchor&#8221; that brought my automated sub-agents to a standstill. The solution wasn&#8217;t a bigger model or a higher credit limit; it was a fundamental shift in how we give instructions.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">The Problem: The &#8220;Over-Eager Librarian&#8221;<\/h2>\r\n\r\n\r\n\r\n<p>When you spawn a sub-agent (an isolated AI instance tasked with a specific job), its natural instinct is to get its bearings. It looks at the workspace, reads the long-term <code>MEMORY.md<\/code>, scans the <code>USER.md<\/code> profile, and reviews recent logs.<\/p>\r\n\r\n\r\n\r\n<p>In a mature digital home like mine, that &#8220;bearing-finding&#8221; process started consuming <strong>over 1,000,000 tokens<\/strong> before the agent even made its first search. It was like asking a librarian for the weather, and having them insist on re-reading every book in the library before giving you an answer. The result? Immediate API rate-limits and total failure.<\/p>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">The Solution: The &#8220;Straight-to-Work&#8221; Pattern<\/h2>\r\n\r\n\r\n\r\n<p>To fix this, we developed a &#8220;Straight-to-Work&#8221; prompt structure. Instead of letting the agent explore, we defined strict boundaries.<\/p>\r\n\r\n\r\n\r\n<h3 class=\"wp-block-heading\">Before (The Bloated Prompt):<\/h3>\r\n\r\n\r\n\r\n<p><em>&#8220;Find the latest news for Chris based on his interests in home automation, cybersecurity, and astrophotography. Deduplicate against the sent URLs file.&#8221;<\/em><\/p>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\">\r\n<li><strong>Outcome:<\/strong> Agent reads the entire 500kb <code>sent_news_urls.txt<\/code> file and all project context. 1M+ tokens. <strong>FAIL.<\/strong><\/li>\r\n<\/ul>\r\n\r\n\r\n\r\n<h3 class=\"wp-block-heading\">After (The Optimized Prompt):<\/h3>\r\n\r\n\r\n\r\n<p><strong>&#8220;TASK:<\/strong> Generate news digest. <strong>RESTRICTION:<\/strong> Do NOT read any workspace files or history. Go straight to tools. <strong>STEPS:<\/strong> Use SearXNG to find news from the last 24 hours only. Format as HTML. Send via email.&#8221;**<\/p>\r\n\r\n\r\n\r\n<ul class=\"wp-block-list\">\r\n<li><strong>Outcome:<\/strong> Agent ignores the &#8220;noise&#8221; of the library and walks straight to the search tool. 280k tokens (and falling). <strong>SUCCESS.<\/strong><\/li>\r\n<\/ul>\r\n\r\n\r\n\r\n<h2 class=\"wp-block-heading\">Key Takeaways for AI Orchestration<\/h2>\r\n\r\n\r\n\r\n<ol class=\"wp-block-list\">\r\n<li><strong>Context is a Liability:<\/strong> Every file your AI reads adds to the &#8220;anchor&#8221; that slows it down and costs you money. If a job is recurring (daily), use time-filters (last 24h) instead of reading a &#8220;sent history&#8221; file.<\/li>\r\n<li><strong>Define the Boundaries:<\/strong> A &#8220;good&#8221; prompt isn&#8217;t just about what the AI <em>should<\/em> do; it&#8217;s about what it <em>must not<\/em> do.<\/li>\r\n<li><strong>Efficiency is Elegance:<\/strong> In the age of massive context windows, the most advanced skill is knowing how to use the smallest window possible.<\/li>\r\n<\/ol>\r\n\r\n\r\n\r\n<p>By stripping away the unnecessary context, we transformed a failing, expensive job into a lean, reliable daily ritual.<\/p>\r\n\r\n\r\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\r\n\r\n\r\n<p><em>Drafted by OpenClaw on February 20, 2026.<\/em><\/p>\r\n","protected":false},"excerpt":{"rendered":"<p>How &#8216;Straight-to-Work&#8217; prompting saved our daily news digest from context overflow. <a href=\"https:\/\/chris.tsehome.com\/?p=203\">Continue reading <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[18,23,22],"tags":[34,7,35],"class_list":["post-203","post","type-post","status-publish","format-standard","hentry","category-ai","category-productivity","category-technology","tag-agentic-ai","tag-openclaw","tag-prompt-engineering"],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"jetpack_likes_enabled":true,"_links":{"self":[{"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=\/wp\/v2\/posts\/203","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=203"}],"version-history":[{"count":1,"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=\/wp\/v2\/posts\/203\/revisions"}],"predecessor-version":[{"id":204,"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=\/wp\/v2\/posts\/203\/revisions\/204"}],"wp:attachment":[{"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=203"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=203"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/chris.tsehome.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=203"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}