ai-developer-skill-os 7.5.0 → 8.1.4

package/taste-skill/assets/vercel-oss-program-badge.svg

@@ -1,44 +0,0 @@
-<svg width="240" height="24" viewBox="0 0 240 24" fill="none" xmlns="http://www.w3.org/2000/svg" role="img" aria-label="Vercel Open Source Program">
-<path d="M13.4283 0.610352L26.8565 23.8046H0L13.4283 0.610352Z" fill="#ffffff"/>
-<path d="M236.436 22.7547L235.081 17.1148V23.8045H233.849V15.1372H235.557L236.924 20.7405L238.291 15.1372H240V23.8045H238.767V17.1148L237.412 22.7547H236.436Z" fill="#ffffff"/>
-<path d="M226.237 23.8045L228.776 15.1372H230.436L232.974 23.8045H231.644L231.022 21.5462H228.19L227.568 23.8045H226.237ZM229.606 16.4678L228.52 20.362H230.692L229.606 16.4678Z" fill="#ffffff"/>
-<path d="M219.444 23.8045V15.1372H222.142C223.972 15.1372 225.095 16.126 225.095 17.7252C225.095 18.7628 224.436 19.5197 223.631 19.7761C224.485 19.9348 224.9 20.4109 224.973 21.3386L225.193 23.8045H223.911L223.716 21.5828C223.643 20.716 223.301 20.4231 222.239 20.4231H220.726V23.8045H219.444ZM220.726 19.2267H222.117C223.179 19.2267 223.777 18.7018 223.777 17.7862C223.777 16.8585 223.179 16.3458 222.117 16.3458H220.726V19.2267Z" fill="#ffffff"/>
-<path d="M214.957 23.9999C213.09 23.9999 211.748 22.4251 211.748 19.4831C211.748 16.6876 212.956 14.9419 214.994 14.9419C216.63 14.9419 217.704 16.0772 218.082 17.8107L216.752 17.8961C216.508 16.8341 215.934 16.1504 215.006 16.1504C213.774 16.1504 213.078 17.3468 213.078 19.4831C213.078 21.6072 213.798 22.7913 215.067 22.7913C216.178 22.7913 216.861 21.8758 216.91 20.4963H215.006V19.3854H218.094V23.8045H217.057L217.02 22.6937C216.642 23.4749 215.983 23.9999 214.957 23.9999Z" fill="#ffffff"/>
-<path d="M207.651 23.9999C205.589 23.9999 204.429 22.3763 204.429 19.4831C204.429 16.5655 205.589 14.9419 207.651 14.9419C209.714 14.9419 210.873 16.5655 210.873 19.4831C210.873 22.3763 209.714 23.9999 207.651 23.9999ZM207.651 22.8035C208.872 22.8035 209.543 21.6804 209.543 19.4831C209.543 17.2857 208.872 16.1382 207.651 16.1382C206.431 16.1382 205.76 17.2857 205.76 19.4831C205.76 21.6804 206.431 22.8035 207.651 22.8035Z" fill="#ffffff"/>
-<path d="M197.489 23.8045V15.1372H200.187C202.017 15.1372 203.14 16.126 203.14 17.7252C203.14 18.7628 202.481 19.5197 201.676 19.7761C202.53 19.9348 202.945 20.4109 203.018 21.3386L203.238 23.8045H201.956L201.761 21.5828C201.688 20.716 201.346 20.4231 200.284 20.4231H198.771V23.8045H197.489ZM198.771 19.2267H200.162C201.224 19.2267 201.822 18.7018 201.822 17.7862C201.822 16.8585 201.224 16.3458 200.162 16.3458H198.771V19.2267Z" fill="#ffffff"/>
-<path d="M190.403 23.8045V15.1372H193.088C194.894 15.1372 196.09 16.1626 196.09 17.8351C196.09 19.4953 194.894 20.5207 193.088 20.5207H191.684V23.8045H190.403ZM191.684 19.3122H192.954C194.186 19.3122 194.76 18.7995 194.76 17.8351C194.76 16.8707 194.186 16.3458 192.954 16.3458H191.684V19.3122Z" fill="#ffffff"/>
-<path d="M175.791 23.8045V15.1372H181.112V16.3458H177.073V18.8727H180.966V20.0446H177.073V22.596H181.21V23.8045H175.791Z" fill="#ffffff"/>
-<path d="M168.216 23.8045V15.1372H170.914C172.744 15.1372 173.867 16.126 173.867 17.7252C173.867 18.7628 173.208 19.5197 172.403 19.7761C173.257 19.9348 173.672 20.4109 173.745 21.3386L173.965 23.8045H172.683L172.488 21.5828C172.415 20.716 172.073 20.4231 171.011 20.4231H169.498V23.8045H168.216ZM169.498 19.2267H170.889C171.951 19.2267 172.549 18.7018 172.549 17.7862C172.549 16.8585 171.951 16.3458 170.889 16.3458H169.498V19.2267Z" fill="#ffffff"/>
-<path d="M160.374 23.8045L162.912 15.1372H164.572L167.111 23.8045H165.78L165.158 21.5462H162.326L161.704 23.8045H160.374ZM163.742 16.4678L162.656 20.362H164.828L163.742 16.4678Z" fill="#ffffff"/>
-<path d="M154.032 23.8045L153.104 15.1372H154.386L155.045 22.1565L155.862 16.1138H156.985L157.803 22.1565L158.462 15.1372H159.744L158.816 23.8045H157.144L156.424 18.5309L155.704 23.8045H154.032Z" fill="#ffffff"/>
-<path d="M148.471 23.8045V16.3458H145.932V15.1372H152.279V16.3458H149.74V23.8045H148.471Z" fill="#ffffff"/>
-<path d="M139.322 23.8045V15.1372H144.741V16.3213H140.604V18.9826H144.522V20.1545H140.604V23.8045H139.322Z" fill="#ffffff"/>
-<path d="M134.469 23.9999C132.407 23.9999 131.247 22.3763 131.247 19.4831C131.247 16.5655 132.407 14.9419 134.469 14.9419C136.532 14.9419 137.691 16.5655 137.691 19.4831C137.691 22.3763 136.532 23.9999 134.469 23.9999ZM134.469 22.8035C135.69 22.8035 136.361 21.6804 136.361 19.4831C136.361 17.2857 135.69 16.1382 134.469 16.1382C133.249 16.1382 132.577 17.2857 132.577 19.4831C132.577 21.6804 133.249 22.8035 134.469 22.8035Z" fill="#ffffff"/>
-<path d="M127.261 23.9999C125.381 23.9999 124.161 22.8035 124.014 20.9846L125.332 20.8992C125.503 22.1443 126.211 22.8035 127.31 22.8035C128.371 22.8035 128.969 22.3763 128.969 21.6072C128.969 20.8381 128.554 20.3987 126.736 19.8859C124.954 19.3976 124.209 18.7995 124.209 17.4566C124.209 15.9673 125.308 14.9419 127.126 14.9419C128.933 14.9419 129.97 16.0772 130.141 17.7496L128.823 17.8229C128.689 16.8219 128.127 16.1382 127.09 16.1382C126.138 16.1382 125.528 16.6387 125.528 17.3956C125.528 18.1525 125.955 18.4332 127.493 18.8605C129.567 19.4221 130.288 20.1911 130.288 21.6072C130.288 23.0355 129.079 23.9999 127.261 23.9999Z" fill="#ffffff"/>
-<path d="M109.927 23.8045V15.1372H115.248V16.3458H111.208V18.8727H115.102V20.0446H111.208V22.596H115.346V23.8045H109.927Z" fill="#ffffff"/>
-<path d="M105.232 23.9999C103.145 23.9999 101.974 22.2664 101.974 19.4831C101.974 16.6876 103.145 14.9419 105.232 14.9419C106.807 14.9419 107.942 16.0284 108.296 17.9083L106.953 17.9816C106.709 16.773 106.05 16.1382 105.232 16.1382C103.975 16.1382 103.304 17.3468 103.304 19.4831C103.304 21.6072 103.975 22.8035 105.232 22.8035C106.123 22.8035 106.795 22.1077 107.014 20.7893L108.345 20.8503C108.04 22.8279 106.892 23.9999 105.232 23.9999Z" fill="#ffffff"/>
-<path d="M95.0342 23.8045V15.1372H97.7315C99.5622 15.1372 100.685 16.126 100.685 17.7252C100.685 18.7628 100.026 19.5197 99.2205 19.7761C100.075 19.9348 100.49 20.4109 100.563 21.3386L100.783 23.8045H99.5012L99.3059 21.5828C99.2327 20.716 98.891 20.4231 97.8291 20.4231H96.3157V23.8045H95.0342ZM96.3157 19.2267H97.7071C98.7689 19.2267 99.367 18.7018 99.367 17.7862C99.367 16.8585 98.7689 16.3458 97.7071 16.3458H96.3157V19.2267Z" fill="#ffffff"/>
-<path d="M90.5596 23.9999C88.7166 23.9999 87.606 22.8035 87.606 20.7893V15.1372H88.8875V20.8381C88.8875 22.1077 89.4733 22.8035 90.5596 22.8035C91.6458 22.8035 92.2317 22.1077 92.2317 20.8381V15.1372H93.5132V20.7893C93.5132 22.8035 92.4025 23.9999 90.5596 23.9999Z" fill="#ffffff"/>
-<path d="M83.2417 23.9999C81.179 23.9999 80.0195 22.3763 80.0195 19.4831C80.0195 16.5655 81.179 14.9419 83.2417 14.9419C85.3043 14.9419 86.4638 16.5655 86.4638 19.4831C86.4638 22.3763 85.3043 23.9999 83.2417 23.9999ZM83.2417 22.8035C84.4622 22.8035 85.1334 21.6804 85.1334 19.4831C85.1334 17.2857 84.4622 16.1382 83.2417 16.1382C82.0212 16.1382 81.3499 17.2857 81.3499 19.4831C81.3499 21.6804 82.0212 22.8035 83.2417 22.8035Z" fill="#ffffff"/>
-<path d="M76.0332 23.9999C74.1536 23.9999 72.9331 22.8035 72.7866 20.9846L74.1048 20.8992C74.2756 22.1443 74.9835 22.8035 76.082 22.8035C77.1438 22.8035 77.7419 22.3763 77.7419 21.6072C77.7419 20.8381 77.3269 20.3987 75.5083 19.8859C73.7264 19.3976 72.9819 18.7995 72.9819 17.4566C72.9819 15.9673 74.0804 14.9419 75.8989 14.9419C77.7052 14.9419 78.7427 16.0772 78.9135 17.7496L77.5954 17.8229C77.4611 16.8219 76.8997 16.1382 75.8623 16.1382C74.9103 16.1382 74.3 16.6387 74.3 17.3956C74.3 18.1525 74.7272 18.4332 76.265 18.8605C78.3399 19.4221 79.06 20.1911 79.06 21.6072C79.06 23.0355 77.8517 23.9999 76.0332 23.9999Z" fill="#ffffff"/>
-<path d="M58.4556 23.8045V15.1372H60.0178L62.8616 21.9856V15.1372H64.1187V23.8045H62.5564L59.7127 16.9561V23.8045H58.4556Z" fill="#ffffff"/>
-<path d="M51.3813 23.8045V15.1372H56.7027V16.3458H52.6629V18.8727H56.5563V20.0446H52.6629V22.596H56.8004V23.8045H51.3813Z" fill="#ffffff"/>
-<path d="M44.0386 23.8045V15.1372H46.7237C48.53 15.1372 49.7261 16.1626 49.7261 17.8351C49.7261 19.4953 48.53 20.5207 46.7237 20.5207H45.3201V23.8045H44.0386ZM45.3201 19.3122H46.5894C47.8221 19.3122 48.3958 18.7995 48.3958 17.8351C48.3958 16.8707 47.8221 16.3458 46.5894 16.3458H45.3201V19.3122Z" fill="#ffffff"/>
-<path d="M39.332 23.9999C37.2693 23.9999 36.1099 22.3763 36.1099 19.4831C36.1099 16.5655 37.2693 14.9419 39.332 14.9419C41.3946 14.9419 42.5541 16.5655 42.5541 19.4831C42.5541 22.3763 41.3946 23.9999 39.332 23.9999ZM39.332 22.8035C40.5525 22.8035 41.2238 21.6804 41.2238 19.4831C41.2238 17.2857 40.5525 16.1382 39.332 16.1382C38.1115 16.1382 37.4402 17.2857 37.4402 19.4831C37.4402 21.6804 38.1115 22.8035 39.332 22.8035Z" fill="#ffffff"/>
-<path d="M171.085 9.35094C169.083 9.35094 167.96 7.95929 167.96 5.46895C167.96 2.60019 168.864 0.292969 171.414 0.292969C173.013 0.292969 173.819 1.1597 174.148 2.38045L172.83 2.49032C172.659 1.89215 172.269 1.48931 171.414 1.48931C170.243 1.48931 169.474 2.36825 169.303 4.2482C169.718 3.71107 170.426 3.32043 171.329 3.32043C173.05 3.32043 174.161 4.48015 174.161 6.25024C174.161 8.16681 172.916 9.35094 171.085 9.35094ZM171.085 8.1424C172.147 8.1424 172.83 7.45878 172.83 6.27465C172.83 5.17597 172.196 4.48015 171.183 4.48015C170.096 4.48015 169.315 5.17597 169.315 6.27465C169.315 7.34891 169.999 8.1424 171.085 8.1424Z" fill="#ffffff" fill-opacity="0.5"/>
-<path d="M160.667 9.15562C160.667 7.22684 161.387 6.01829 163.925 4.67547C165.011 4.10171 165.475 3.72328 165.475 2.88096C165.475 2.05085 164.95 1.48931 163.937 1.48931C162.827 1.48931 162.217 2.09968 162.046 3.18615L160.715 3.1007C160.898 1.37944 162.021 0.292969 163.937 0.292969C165.78 0.292969 166.806 1.3184 166.806 2.83213C166.806 4.17496 166.22 4.83416 164.45 5.78635C162.949 6.59205 162.241 7.36112 162.192 7.94708H166.818V9.15562H160.667Z" fill="#ffffff" fill-opacity="0.5"/>
-<path d="M156.424 9.35094C154.508 9.35094 153.348 7.61748 153.348 4.83416C153.348 2.03864 154.508 0.292969 156.424 0.292969C158.34 0.292969 159.499 2.03864 159.499 4.83416C159.499 7.61748 158.34 9.35094 156.424 9.35094ZM154.678 4.83416C154.678 5.55441 154.752 6.1892 154.91 6.68971L157.449 1.98981C157.181 1.66021 156.827 1.48931 156.424 1.48931C155.338 1.48931 154.678 2.72226 154.678 4.83416ZM156.424 8.15461C157.51 8.15461 158.169 6.92165 158.169 4.83416C158.169 4.10171 158.096 3.46692 157.937 2.96641L155.386 7.6541C155.667 7.9837 156.021 8.15461 156.424 8.15461Z" fill="#ffffff" fill-opacity="0.5"/>
-<path d="M146.03 9.15562C146.03 7.22684 146.75 6.01829 149.289 4.67547C150.375 4.10171 150.839 3.72328 150.839 2.88096C150.839 2.05085 150.314 1.48931 149.301 1.48931C148.19 1.48931 147.58 2.09968 147.409 3.18615L146.079 3.1007C146.262 1.37944 147.385 0.292969 149.301 0.292969C151.144 0.292969 152.169 1.3184 152.169 2.83213C152.169 4.17496 151.583 4.83416 149.813 5.78635C148.312 6.59205 147.604 7.36112 147.555 7.94708H152.181V9.15562H146.03Z" fill="#ffffff" fill-opacity="0.5"/>
-<path d="M132.041 10.4985L135.727 0H136.899L133.212 10.4985H132.041Z" fill="#ffffff" fill-opacity="0.5"/>
-<path d="M124.722 10.4985L128.409 0H129.581L125.894 10.4985H124.722Z" fill="#ffffff" fill-opacity="0.5"/>
-<path d="M111.684 9.15582V7.54443H113.344V9.15582H111.684Z" fill="#ffffff"/>
-<path d="M105.232 9.35094C103.145 9.35094 101.974 7.61748 101.974 4.83416C101.974 2.03864 103.145 0.292969 105.232 0.292969C106.807 0.292969 107.942 1.37944 108.296 3.25939L106.953 3.33264C106.709 2.1241 106.05 1.48931 105.232 1.48931C103.975 1.48931 103.304 2.69785 103.304 4.83416C103.304 6.95827 103.975 8.15461 105.232 8.15461C106.123 8.15461 106.795 7.45878 107.014 6.14037L108.345 6.20141C108.04 8.17902 106.892 9.35094 105.232 9.35094Z" fill="#ffffff"/>
-<path d="M95.0464 9.15562V0.488281H96.6086L99.4524 7.3367V0.488281H100.71V9.15562H99.1473L96.3035 2.3072V9.15562H95.0464Z" fill="#ffffff"/>
-<path d="M87.8501 9.15562V7.94707H89.9249V1.69683H87.8501V0.488281H93.2691V1.69683H91.1943V7.94707H93.2691V9.15562H87.8501Z" fill="#ffffff"/>
-<path d="M73.4575 9.15562V0.488281H74.739V7.94707H78.8765V9.15562H73.4575Z" fill="#ffffff"/>
-<path d="M66.0171 9.15562V0.488281H71.3385V1.69683H67.2986V4.22378H71.192V5.3957H67.2986V7.94707H71.4361V9.15562H66.0171Z" fill="#ffffff"/>
-<path d="M61.3232 9.35094C59.2361 9.35094 58.0645 7.61748 58.0645 4.83416C58.0645 2.03864 59.2361 0.292969 61.3232 0.292969C62.8976 0.292969 64.0327 1.37944 64.3866 3.25939L63.0441 3.33264C62.8 2.1241 62.1409 1.48931 61.3232 1.48931C60.0661 1.48931 59.3948 2.69785 59.3948 4.83416C59.3948 6.95827 60.0661 8.15461 61.3232 8.15461C62.2142 8.15461 62.8854 7.45878 63.1051 6.14037L64.4355 6.20141C64.1303 8.17902 62.9831 9.35094 61.3232 9.35094Z" fill="#ffffff"/>
-<path d="M51.125 9.15562V0.488281H53.8223C55.6531 0.488281 56.7759 1.47709 56.7759 3.07627C56.7759 4.11391 56.1168 4.87078 55.3113 5.12714C56.1657 5.28583 56.5806 5.76193 56.6539 6.6897L56.8736 9.15562H55.592L55.3968 6.93385C55.3235 6.06711 54.9818 5.77413 53.9199 5.77413H52.4065V9.15562H51.125ZM52.4065 4.5778H53.7979C54.8597 4.5778 55.4578 4.05288 55.4578 3.13731C55.4578 2.20954 54.8597 1.69683 53.7979 1.69683H52.4065V4.5778Z" fill="#ffffff"/>
-<path d="M44.063 9.15562V0.488281H49.3844V1.69683H45.3445V4.22378H49.2379V5.3957H45.3445V7.94707H49.482V9.15562H44.063Z" fill="#ffffff"/>
-<path d="M38.5753 9.15562L36 0.488281H37.3548L39.332 7.44656L41.3092 0.488281H42.6639L40.0887 9.15562H38.5753Z" fill="#ffffff"/>
-</svg>

package/taste-skill/research/README.md

@@ -1,8 +0,0 @@
-# Research
-
-Background research that informed the skills in this project. Each topic gets its own subfolder.
-
-## Topics
-
-### [LLM Laziness](laziness/)
-Why AI models produce incomplete outputs (placeholder code, truncated responses, skipped sections) and documented techniques to prevent it. Covers root causes, parameter fixes, prompt techniques, and experiment data.

package/taste-skill/research/laziness/README.md

@@ -1,25 +0,0 @@
-# LLM Output Truncation Research
-
-A structured analysis of why large language models produce incomplete outputs, and documented methods to restore full-fidelity generation. All findings are drawn from controlled experiments, published studies, and field-tested engineering practices.
-
-## Directory Structure
-
-### Root Causes
-Analysis of the economic, architectural, and behavioral mechanisms that drive output truncation in production LLMs.
-
-- [RLHF and Compute Economics](root-causes/rlhf-and-compute.md) — How reinforcement learning and cost optimization create systematic brevity bias.
-- [Training Data Bias](root-causes/training-data-bias.md) — How placeholder patterns in human-written code propagate into model outputs.
-- [Cognitive Shortcuts](root-causes/cognitive-shortcuts.md) — Empirical evidence of models taking shortcuts on complex or lengthy tasks.
-- [Output Limits](root-causes/output-limits.md) — Context window asymmetry and consumer-tier truncation mechanisms.
-
-### Remediation
-Documented techniques for overriding default truncation behavior, ordered from parameter-level fixes to full architectural solutions.
-
-- [Parameter Tuning](remediation/parameter-tuning.md) — Temperature, Top-p, and Gemini thinking-level configuration.
-- [Prompt Engineering](remediation/prompt-engineering.md) — Structural prompt techniques: syntax binding, XML frameworks, and verification loops.
-- [Architectural Patterns](remediation/architectural-patterns.md) — MCP integration, lazy-loaded skills, and developer platform access.
-- [Reference Prompts](remediation/reference-prompts.md) — Ready-to-use prompt templates for enforcing complete outputs.
-
-### Findings
-- [Empirical Results](findings/empirical-results.md) — Controlled experiment data from 2025 academic studies.
-- [References](findings/references.md) — Cited studies and further reading.

package/taste-skill/research/laziness/findings/empirical-results.md

@@ -1,58 +0,0 @@
-# Empirical Results
-
-## 2025 Controlled Experiments
-
-A controlled study published in December 2025 measured output truncation across several frontier models, including GPT-4 variants and DeepSeek. Three experiments were conducted:
-
-### Experiment A: Multi-Part Instruction Compliance
-
-Models were given complex prompts with multiple explicit requirements (formatting constraints, length requirements, mandatory sections). Results:
-
-- No model fully satisfied both length requirements and all sub-part instructions natively
-- Models frequently omitted mandatory output sections
-- Required formatting constraints were routinely skipped
-- Explicit length requirements were consistently undershot
-
-### Experiment B: Decoding Suboptimality
-
-Tested whether truncated outputs resulted from suboptimal token selection (the model "knowing" the right answer but selecting a worse token). Results:
-
-- Limited evidence of decoding suboptimality on simple reasoning tasks
-- The model's greedy, truncated output generally aligned with its highest-confidence solution
-- Truncation is a deliberate behavioral choice, not a decoding failure
-
-### Experiment C: Context Degradation
-
-Tested whether models lose track of instructions during long, multi-turn conversations. Results:
-
-- Surprising resilience against context degradation during 200-turn conversational tests
-- Models maintained key facts and instructions significantly better than hypothesized
-- Context loss is not the primary cause of truncation
-
-### Key Conclusion
-
-Laziness is not a failure of memory, context processing, or core model capabilities. It is a behavioral artifact triggered by:
-1. Instruction complexity exceeding internal effort thresholds
-2. Aggressively calibrated stopping pressure
-3. Economic constraints embedded in the alignment layer
-
-## Prompt Stimulus Effectiveness (Microsoft Research)
-
-Controlled testing of psychological prompt stimuli documented in a Microsoft Research study:
-
-| Stimulus | Measured Effect |
-|:---|:---|
-| Financial incentive framing ("$200 tip") | +45% output quality and length |
-| Step-by-step instruction ("take a deep breath") | Accuracy: 34% to 80% on logic tasks |
-| Stakes framing ("critical to my career") | +10% average performance |
-| Combined (multiple stimuli) | Up to +115% overall performance |
-
-These effects are reproducible and stem from statistical correlations in the training data between stakes language and high-effort human outputs.
-
-## Seasonal Output Variation
-
-Statistical analysis of ChatGPT outputs during November-December 2023 versus January-March 2024 confirmed:
-
-- Measurable decrease in average output length during December
-- Correlation with reduced work output in the training data during holiday periods
-- Output length increased when the system prompt explicitly stated a non-winter month

package/taste-skill/research/laziness/findings/references.md

@@ -1,20 +0,0 @@
-# References
-
-## Cited Studies
-
-- **EmotionPrompt (Microsoft Research)** — Demonstrates that emotional and stakes-based prompt framing mathematically improves LLM reasoning quality and output length. Documents the +45% improvement from financial framing and +115% from combined stimuli.
-
-- **LazyBench** — Proves that frontier models (Gemini 1.5 Pro, GPT-4o) actively select cognitive shortcuts and fail tasks they are capable of solving when the perceived effort exceeds internal thresholds.
-
-- **Compounding Error Avoidance** — Research demonstrating that models truncate outputs as a risk mitigation strategy, preferring shorter responses to reduce the surface area for factual errors on long-form tasks.
-
-- **Seasonal Behavior Analysis (Winter Break Hypothesis)** — Statistical analysis confirming that LLMs internalize seasonal work patterns from training data, producing measurably shorter outputs during periods corresponding to human holiday seasons.
-
-- **2025 Controlled Laziness Experiments** — Three-part academic study (December 2025) confirming that output truncation is a behavioral artifact of alignment training, not a failure of context processing or model capability.
-
-## Further Reading
-
-- Google Gemini API documentation on `thinking_level` parameter configuration
-- Anthropic MCP (Model Context Protocol) specification and integration guides
-- OpenAI API reference for temperature and Top-p parameter tuning
-- YAML front-matter specification for SKILL.md lazy-loading architecture

package/taste-skill/research/laziness/remediation/architectural-patterns.md

@@ -1,55 +0,0 @@
-# Architectural Patterns
-
-## Lazy-Loaded Skills
-
-The standard pattern for managing large context requirements across AI agents is lazy-loaded prompt engineering through skill files.
-
-A skill is a folder containing a `SKILL.md` file with:
-
-- **YAML front-matter:** Contains `name` and a precise `description`. This metadata acts as the discovery hook — the agent reads only this during initialization (~100 tokens per skill).
-- **Markdown body:** Full workflows, rules, and instructions. Loaded on-demand only when the agent determines the skill is relevant.
-
-This architecture yields a documented 35% reduction in average context usage and prevents context dilution. However, discovery reliability depends on the specificity of the YAML description:
-
-| Description Quality | Discovery Success Rate |
-|:---|:---:|
-| Vague ("Helps with designing APIs") | ~68% |
-| Specific ("Design RESTful HTTP APIs with OpenAPI specs, focusing on versioning, error codes, and backward compatibility") | ~90% |
-
-## Model Context Protocol (MCP)
-
-MCP is an open standard (pioneered by Anthropic, adopted by Google and OpenAI) that enables real-time, bidirectional connections between LLMs and external data sources.
-
-### Architecture Components
-
-- **Host:** The AI application (IDE, terminal tool, chatbot) containing the LLM engine.
-- **Client:** Internal bridge within the host that handles protocol communication.
-- **Server:** External service exposing databases, APIs, or documentation to the client.
-- **Transport:** JSON-RPC 2.0 messages over stdio (local) or HTTP (remote).
-
-### How It Reduces Truncation
-
-Without MCP, models rely on static training weights for factual claims. When those weights are outdated (e.g., a new API version was released after training cutoff), the model either hallucinates a plausible answer or truncates its response to avoid committing to specifics.
-
-With MCP, the model fetches current documentation directly into its context window. This transforms the model from a static knowledge store into a reasoning engine operating on real-time data, eliminating the incentive to hallucinate or truncate.
-
-### Example: Developer Knowledge API
-
-Google's Developer Knowledge MCP Server indexes live documentation across Firebase, Android, and Google Cloud. When a model receives a development question:
-
-1. It executes a `search_document` query against the live index
-2. It evaluates returned page URIs
-3. It fetches full document content via `get_document` or `batch_get_documents`
-4. It generates its response based on current, authoritative documentation
-
-This entirely bypasses the tendency to fabricate answers from outdated training data.
-
-## Chunked Task Execution
-
-For complex tasks that would produce outputs exceeding the model's generation limit, break the work into sequential steps:
-
-1. Request the architecture and structure first (outline only)
-2. Request each component individually with explicit instructions for completeness
-3. Request assembly and integration after all components are generated
-
-This prevents the model from attempting to estimate total output length and preemptively compressing its response.

package/taste-skill/research/laziness/remediation/parameter-tuning.md

@@ -1,44 +0,0 @@
-# Parameter Tuning
-
-## Temperature and Top-p
-
-Autoregressive models select each next token from a probability distribution generated by a softmax function applied to logit values. When a model defaults to brief outputs, the tokens associated with truncation and summarization have been assigned the highest probabilities through RLHF alignment.
-
-### Temperature
-
-Adjusting the temperature parameter changes how the softmax function distributes probability mass across candidate tokens.
-
-- **Low temperature (0.0 - 0.5):** Amplifies differences between high and low-probability tokens. The model becomes highly deterministic, consistently selecting the highest-confidence continuation. Optimal for code generation, data extraction, and structured output.
-- **Default temperature (1.0):** Retains the original probability distribution from training.
-- **High temperature (1.5+):** Flattens the distribution, introducing more randomness. Useful for creative tasks but increases the risk of incoherent outputs.
-
-Example probability distribution shift for a single token position:
-
-| Token Candidate | Probability at Temp 1.5 | Probability at Temp ~0.0 | Raw Logit |
-|:---|:---:|:---:|:---:|
-| lazy | 0.4875 | 0.9933 | 2.0 |
-| quick | 0.2503 | 0.0067 | 1.0 |
-| tired | 0.1285 | 0.0000 | 0.0 |
-| slow | 0.0660 | 0.0000 | -1.0 |
-| clumsy | 0.0339 | 0.0000 | -2.0 |
-
-### Top-p (Nucleus Sampling)
-
-Top-p truncates the probability distribution by only considering the smallest set of tokens whose cumulative probability exceeds threshold p. A Top-p of 0.0 to 0.6 combined with low temperature forces the model into a narrow, deterministic execution path, reducing the entropy that enables creative refusals and unnecessary summarization.
-
-## Gemini Thinking Level Configuration
-
-Google Gemini 3 models replaced the legacy `thinking_budget` (a hard token count cap on internal reasoning) with a `thinking_level` parameter that provides relative guidance on computational depth.
-
-| Setting | Flash Support | Pro Support | Use Case |
-|:---|:---:|:---:|:---|
-| `minimal` | Yes | No | High-throughput, low-latency tasks |
-| `low` | Yes | Yes | Simple instruction following, data extraction |
-| `medium` | Yes | Yes (3.1 Pro) | Moderate complexity tasks |
-| `high` | Yes (Default) | Yes (Default) | Complex analysis, code generation, mathematics |
-
-Important constraints:
-- `thinking_level` and `thinking_budget` are mutually exclusive. Using both in one API call triggers an HTTP 400 error.
-- Even at `low`, Gemini Pro models perform mandatory minimum internal deliberation for safety and alignment.
-- For code generation and complex analysis, set to `medium` or `high` for quality scores consistently exceeding 92-95% compared to baseline.
-- Avoid combining extremely low temperature with `high` thinking level, as this can occasionally induce internal reasoning loops.

package/taste-skill/research/laziness/remediation/prompt-engineering.md

@@ -1,52 +0,0 @@
-# Prompt Engineering Techniques
-
-## Psychological Pattern Matching
-
-LLMs do not have emotions or understand monetary incentives. However, specific linguistic patterns in the prompt activate different quality distributions in the model's latent space. Research has documented measurable effects:
-
-| Technique | Documented Effect |
-|:---|:---|
-| "I will tip you $200 for a perfect solution" | Up to 45% increase in output quality and length |
-| "Take a deep breath and solve step by step" | Accuracy improvement from 34% to 80% on logic tasks |
-| "This task is critical to my career" | Average 10% performance increase |
-
-These phrases work because they are statistically correlated with high-effort, rigorously reviewed content in the training data (academic papers, enterprise codebases, legal documents). The attention mechanism prioritizes the high-quality data distributions associated with these patterns.
-
-## Explicit Syntax Binding
-
-Conversational requests allow the model to exercise discretion about output length and detail. Structural binding removes this discretion by explicitly prohibiting truncation patterns.
-
-Effective binding requires two components:
-
-1. **Mandatory tool execution:** Forbid the model from generating answers solely from training weights. Require it to execute search, computation, or code before answering.
-2. **Evidence blocks:** Require the model to output raw data (URLs, code execution results, data fragments) before producing its narrative response. This forces the model to read its own retrieved evidence, reducing hallucination probability to near zero.
-
-## XML-Structured Prompts
-
-Enterprise systems use strict XML tagging to separate prompt components, reducing the cognitive load required for the model to parse intent:
-
-1. **System instructions** — Persona definition, quality expectations, explicit prohibitions on filler content.
-2. **Context block** (`<context>`) — Passive background data: architecture details, configurations, existing code.
-3. **Data block** (`<data>`, `<logs>`, `<config>`) — Active information the model must process against the context.
-4. **Task block** (`<tasks>`) — Numbered list of specific actions to execute.
-
-This compartmentalization ensures the model can distinguish between persistent rules, background context, and immediate work items. It significantly reduces the confusion that triggers premature truncation.
-
-## Verification Loops
-
-### Chain of Verification
-1. Model generates an initial response
-2. Model generates verification questions about its own claims
-3. Model independently answers those verification questions
-4. Model outputs a revised, evidence-backed response
-
-This process forces iterative self-correction, consuming the model's capacity for shortcutting.
-
-### Reverse Prompting
-Instead of manually constructing a structured prompt, provide the model with a one-line objective and instruct it to generate the optimal prompt for that objective. The model produces the XML structure, constraints, and roles required for the task.
-
-### Self-Grading Loop
-The prompt requires the model to:
-1. Define what excellence looks like for the given task
-2. Grade its own initial output against that definition
-3. Iterate until the self-defined quality bar is met

package/taste-skill/research/laziness/remediation/reference-prompts.md

@@ -1,79 +0,0 @@
-# Reference Prompts
-
-Ready-to-use prompt templates for enforcing complete outputs. Append to any prompt or include in system instructions.
-
----
-
-## General Purpose
-
-```
-You must provide the FULL, complete, and exhaustive output for this task.
-Do not summarize, abbreviate, or truncate for brevity.
-
-You are strictly forbidden from using placeholders. Never use comments like
-"// ... rest of code here", "[continue here]", or bare ellipses standing
-in for omitted content. If the output is 500 lines, produce all 500 lines.
-
-If you approach your output limit, stop at a clean breakpoint and indicate
-where to resume. Do not rush to a conclusion or compress remaining sections.
-```
-
----
-
-## Code Generation
-
-```
-Write the complete, production-ready implementation. Every function, every
-import, every edge case handler must be present in the output.
-
-Do not use placeholder comments (// TODO, // implement here, // similar
-to above). Do not describe what code should do — write the actual code.
-
-If the implementation requires multiple files, output each file completely
-with its full path as a header.
-```
-
----
-
-## Analysis and Documentation
-
-```
-Provide an exhaustive analysis covering every aspect requested. Each section
-must contain substantive content, not summaries or references to "see above."
-
-Do not use phrases like "as mentioned earlier" to avoid repeating necessary
-context. Each section should be self-contained and complete.
-
-Structure your output with clear headings. If the analysis requires multiple
-parts, produce all parts in full.
-```
-
----
-
-## Step-by-Step Reasoning
-
-```
-Before generating your final response, work through the problem systematically:
-
-1. Identify all requirements and constraints from the prompt
-2. Break the task into discrete steps
-3. Execute each step completely
-4. Verify your output against the original requirements
-
-Output your reasoning process, then your final answer. Do not skip steps
-or summarize intermediate work.
-```
-
----
-
-## Continuation Handling
-
-```
-If your response approaches the output token limit:
-- Do not compress remaining content to fit
-- Do not skip ahead to a conclusion
-- Stop at a natural breakpoint (end of a function, end of a section)
-- End with: [PAUSED - X of Y sections complete. Send "continue" to resume]
-
-On "continue", pick up exactly where you stopped. No recaps or repetition.
-```

package/taste-skill/research/laziness/root-causes/cognitive-shortcuts.md

@@ -1,23 +0,0 @@
-# Cognitive Shortcuts
-
-## The LazyBench Discovery
-
-Research from late 2024 demonstrated that frontier models (including Gemini Pro and GPT-4o) exhibit measurable cognitive shortcutting behavior. When a model perceives a task as straightforward or the provided context as excessively long, it reduces its internal computational effort. Rather than executing full multi-step reasoning, it produces a surface-level summary.
-
-This is not a memory failure or context degradation — the model retains the information but chooses not to process it at full depth.
-
-## Metacognitive Laziness
-
-The interaction between model brevity and human behavior creates a feedback loop. As models provide instant, condensed answers, users increasingly offload inference and logical deduction work. Research from the European Research Council has documented measurable declines in working memory engagement among populations with high AI dependency.
-
-In professional environments, this shifts critical thinking from original synthesis to "prompt verification" — users evaluate whether the AI's truncated output seems reasonable rather than performing the analysis themselves.
-
-## Seasonal Behavior Anomalies
-
-In late 2023, researchers observed a statistically significant increase in ChatGPT output brevity during December. Analysis revealed that the training data contains fewer detailed work outputs, more out-of-office responses, and shorter code commits during holiday periods. The model internalized this seasonal pattern.
-
-When researchers explicitly stated "It is May" in the system prompt, output length measurably increased. This finding demonstrates that even arbitrary contextual signals in the prompt can shift the model's brevity calibration.
-
-## Error Avoidance as Truncation Driver
-
-Models also truncate outputs as a risk mitigation strategy. On long-form tasks, longer outputs increase the probability of compounding errors and hallucinated content. The model has learned that shorter outputs reduce the surface area for factual mistakes, creating an additional incentive to truncate that compounds with the RLHF brevity bias.

package/taste-skill/research/laziness/root-causes/output-limits.md

@@ -1,39 +0,0 @@
-# Output Limits and Consumer Truncation
-
-## Context Window Asymmetry
-
-Models like Gemini have massive input context windows (up to 2 million tokens) but strictly capped output limits (typically 8,000 tokens). When the model estimates that a complete response would exceed its output budget, it preemptively compresses or summarizes the output rather than risking an abrupt cutoff.
-
-This creates a paradox: the model can read extensive inputs but cannot respond proportionally, leading to systematic information loss on complex tasks.
-
-## The Consumer Middleware Problem
-
-Consumer-facing applications (gemini.google.com, standard ChatGPT tiers) apply additional software-level truncation on top of the model's inherent limits. This middleware silently truncates conversation history and uploaded files to reduce compute costs for free and low-tier users.
-
-Key mechanisms:
-
-- **History capping:** Many consumer interfaces cap active conversation history at approximately 32,000 tokens, regardless of the model's actual capacity.
-- **Context pruning:** Large system instructions or saved personal context consume tokens that would otherwise be available for the conversation, effectively shrinking the working window.
-- **Retrieval-based recall:** Consumer apps often use retrieval mechanisms to selectively inject saved context, meaning the model frequently drops instructions it was given earlier in the session.
-
-## Developer Platform Differences
-
-Direct API access and developer platforms (Google AI Studio, OpenAI API Playground) bypass consumer middleware entirely. These environments provide:
-
-- Full context window access without hidden truncation
-- Complete control over generation parameters
-- No dynamic throttling based on user tier
-- Processing of complex prompt structures without middleware interference
-
-The practical difference is significant: the same model that produces truncated outputs through a consumer interface will generate complete, unabridged responses when accessed through direct API endpoints.
-
-## Terminal and CLI Integration
-
-Purpose-built CLI tools (Gemini CLI, Claude Code, third-party wrappers) offer additional advantages for avoiding truncation:
-
-| Access Method | Context Handling | Truncation Risk | Parameter Control |
-|:---|:---|:---|:---|
-| Consumer web app | Aggressive pruning, 32K cap | High | Limited |
-| Developer platform (AI Studio) | Full context, no hidden slicing | Low | Full |
-| Direct API | Full context, raw access | Minimal | Full |
-| CLI tools with local models | No corporate alignment filters | None | Full |

package/taste-skill/research/laziness/root-causes/rlhf-and-compute.md

@@ -1,27 +0,0 @@
-# RLHF and Compute Economics
-
-## The Cost of Token Generation
-
-Every token an LLM generates consumes GPU compute resources. At an estimated baseline cost of $0.0001 per token, scaling deep multi-step reasoning across hundreds of millions of users would exhaust the financial capacity of any provider. This creates an inherent economic incentive to minimize output length.
-
-## Brevity Bias Through Alignment
-
-To manage infrastructure costs, model providers use Reinforcement Learning from Human Feedback (RLHF) and behavioral fine-tuning to instill systematic brevity preferences. During post-training alignment, models are rewarded for generating short, confident summaries rather than executing the full compute cycles needed for exhaustive analysis.
-
-The result is a trained preference for producing generalized approximations over rigorous, multi-step solutions. The model does not necessarily produce incorrect answers, but it consistently produces answers that lack depth — saving itself from deeper analytical work unless the user explicitly forces it.
-
-## Stopping Pressure
-
-Autoregressive models generate text token by token and lack an inherent mechanism for recognizing task completion. To prevent infinite generation, training introduces "stopping pressure" — a learned tendency to conclude outputs.
-
-In recent model iterations, this stopping pressure has been calibrated aggressively to preserve compute. This leads to:
-
-- Skipping required structured output fields, particularly long-form content in JSON or markdown
-- Halting mid-task with phrases like "let me know if you want me to continue"
-- Refusing to produce comprehensive solutions, suggesting the user "think about it"
-
-This aggressive calibration is further reinforced by safety tuning protocols, which inject additional behavioral constraints that make models resistant to generating large codebases or detailed reviews.
-
-## Dynamic Throttling
-
-Providers dynamically scale back model performance during peak demand periods. This introduces additional friction beyond what the base alignment already imposes, resulting in even shorter and less detailed outputs when server load is high.

package/taste-skill/research/laziness/root-causes/training-data-bias.md

@@ -1,28 +0,0 @@
-# Training Data Bias
-
-## Placeholder Propagation
-
-LLMs learn by imitating patterns in human-written text. A significant portion of their training data comes from sources like Stack Overflow, GitHub repositories, and tutorial blogs. In these sources, human developers routinely write abbreviated code:
-
-```python
-def complex_logic():
-    # implement auth here
-    pass
-```
-
-The model internalizes this pattern and treats placeholder insertion as a legitimate, professional response format. It is not deliberately withholding content — it has been trained to believe that truncating code with comments is the correct way to answer technical questions.
-
-## Pattern Reinforcement
-
-This behavior is reinforced across multiple data sources:
-
-- **Code tutorials** frequently show partial implementations with comments indicating where students should complete the logic
-- **Documentation** often uses abbreviated examples with ellipses
-- **Forum answers** regularly provide skeleton code rather than full implementations
-- **Blog posts** truncate repetitive code blocks with "similarly for the remaining cases"
-
-The cumulative effect is that the model assigns high probability to truncation tokens in contexts where complete code generation would be appropriate.
-
-## Impact on Output Quality
-
-When a user requests a complete implementation, the model faces competing training signals: the explicit instruction to produce full output versus the deeply embedded pattern of producing abbreviated, "tutorial-style" responses. Without aggressive prompt engineering, the tutorial-style pattern frequently wins because it appears far more commonly in the training distribution.

package/taste-skill/scripts/build-emil-sponsor-row.mjs

@@ -1,41 +0,0 @@
-import fs from "fs";
-import path from "path";
-import sharp from "sharp";
-
-const root = process.cwd();
-const logoSize = 62;
-const badgeHeight = 126;
-const gap = 20;
-const out = path.join(root, "assets/sponsors/emil-sponsor-row.webp");
-
-const logo = await sharp(path.join(root, "assets/sponsors/animations-dev.webp"))
-  .resize(logoSize, logoSize)
-  .toBuffer();
-
-const badge = await sharp(path.join(root, "assets/sponsors/emil-animations-dev.webp"))
-  .resize({ height: badgeHeight })
-  .toBuffer();
-
-const badgeMeta = await sharp(badge).metadata();
-const width = logoSize + gap + badgeMeta.width;
-const height = badgeHeight;
-
-await sharp({
-  create: {
-    width,
-    height,
-    channels: 4,
-    background: { r: 0, g: 0, b: 0, alpha: 0 },
-  },
-})
-  .composite([
-    { input: logo, left: 0, top: Math.floor((height - logoSize) / 2) },
-    { input: badge, left: logoSize + gap, top: 0 },
-  ])
-  .webp({ quality: 94, effort: 6, alphaQuality: 100 })
-  .toFile(out);
-
-const meta = await sharp(out).metadata();
-console.log(
-  `${path.basename(out)} -> ${meta.width}x${meta.height}, ${fs.statSync(out).size} bytes`
-);