{"id":1176927,"date":"2026-06-30T09:50:02","date_gmt":"2026-06-30T16:50:02","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?p=1176927"},"modified":"2026-06-30T09:50:04","modified_gmt":"2026-06-30T16:50:04","slug":"skillopt-agent-skills-as-trainable-parameters","status":"publish","type":"post","link":"https:\/\/www.microsoft.com\/en-us\/research\/blog\/skillopt-agent-skills-as-trainable-parameters\/","title":{"rendered":"SkillOpt: Agent skills as trainable parameters"},"content":{"rendered":"\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"2560\" height=\"1441\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-scaled.jpg\" alt=\"SkillOpt blog | three white line icons on an abstract green background | shield icon, gear icon, circle with checkmark icon\" class=\"wp-image-1176928\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-scaled.jpg 2560w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1536x865.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-2048x1153.jpg 2048w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1920x1080.jpg 1920w\" sizes=\"auto, (max-width: 2560px) 100vw, 2560px\" \/><\/figure>\n\n\n\n<div style=\"padding-bottom:0;padding-top:0\" class=\"wp-block-msr-immersive-section alignfull row\">\n\t\n\t<div class=\"container\">\n\t\t<div class=\"wp-block-msr-immersive-section__inner wp-block-msr-immersive-section__inner--narrow\">\n\t\t\t<div class=\"wp-block-columns mb-10 pb-1 pr-1 is-layout-flex wp-container-core-columns-is-layout-8f761849 wp-block-columns-is-layout-flex\" style=\"box-shadow:var(--wp--preset--shadow--outlined)\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<h2 id=\"at-a-glance\" class=\"wp-block-heading h3\">At a glance<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>AI agents often fail because their instructions, or skills, are manually modified with no guarantee of improvement. SkillOpt turns skill editing into a training process, making agent behavior more reliable without changing model weights.<\/li>\n\n\n\n<li>SkillOpt treats an agent skill file as a trainable parameter outside a frozen target model, turning skill writing from one-shot prompting into a controlled optimization process.<\/li>\n\n\n\n<li>Across six benchmarks, seven target models, and three execution modes, SkillOpt is the best or tied-best method in all 52 evaluation cells, improving performance without updating model weights.<\/li>\n\n\n\n<li>SkillOpt keeps skills compact and auditable through bounded text edits, validation gating, rejected-edit feedback, and slow\/meta updates, avoiding uncontrolled prompt drift.<\/li>\n\n\n\n<li>The optimized skills transfer across model scales, agent harnesses, and related tasks, suggesting that they capture reusable workflow knowledge rather than benchmark-specific instructions.<\/li>\n<\/ul>\n<\/div>\n<\/div>\t\t<\/div>\n\t<\/div>\n\n\t<\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Large language models (LLMs) are increasingly deployed as agents that gather evidence, call tools, and execute multi-step tasks. For these agents, the hard problem is no longer whether they can call a tool, but whether they can complete tasks reliably and consistently. Today, agent skills typically come from three sources: experts write them by hand, a frontier model generates them one-shot, or the agent loosely revises them after execution. None of these approaches behaves like a deep-learning optimizer. They lack step-size control, held-out validation, and any memory of revisions that failed. As a result, skills tend to grow longer and drift with each rewrite, and a revision that seems perfectly reasonable can quietly degrade real task performance. This uncontrolled skill evolution has become a major obstacle on the path from agent prototype to dependable, production-grade deployment.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In our recent paper, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/skillopt-executive-strategy-for-self-evolving-agent-skills\/\" type=\"link\" id=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/skillopt-executive-strategy-for-self-evolving-agent-skills\/\">SkillOpt: Executive Strategy for Self-Evolving Agent Skills<\/a>, we reframe the question from \u201chow do we write a better prompt?\u201d to \u201chow do we train the skill?\u201d SkillOpt treats the skill file as a trainable parameter living outside a frozen target model, bringing a training-style optimization loop, consistent gains across 52 evaluation cells, and a compact skill file that stays readable, auditable, and transferable.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"586\" height=\"291\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-1.png\" alt=\"Figure 1. A frozen target model executes tasks while a separate optimizer model trains the skill layer from trajectory feedback, exporting the reusable skill file best_ skill.md through validation gating.\" class=\"wp-image-1176930\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-1.png 586w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-1-300x149.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-1-240x119.png 240w\" sizes=\"auto, (max-width: 586px) 100vw, 586px\" \/><figcaption class=\"wp-element-caption\">Figure 1. A frozen target model executes tasks while a separate optimizer model trains the skill layer from trajectory feedback, exporting the reusable skill file best_ skill.md through validation gating.<\/figcaption><\/figure>\n\n\n\n<h2 id=\"how-skillopt-works\" class=\"wp-block-heading\">How SkillOpt works<\/h2>\n\n\n\n<figure class=\"wp-block-video aligncenter\"><video height=\"1080\" style=\"aspect-ratio: 1920 \/ 1080;\" width=\"1920\" controls poster=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/skillopt.jpg\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/skillopt_teaser-1.mp4\" playsinline><\/video><figcaption class=\"wp-element-caption\">Video 1. SkillOpt\u2019s optimization loop, from trajectory collection to the exported skill file.<\/figcaption><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">SkillOpt organizes skill editing as a forward\u2013backward\u2013update cycle in text space. In the forward pass, the frozen target model executes a batch of training tasks with the current skill; the rollout batch size controls how much evidence each update receives. In the backward pass, a separate optimizer model reads the resulting trajectories in reflection minibatches, distilling patterns to preserve from successful trajectories and patterns to correct from failures.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In the update step, the optimizer proposes small add, delete, and replace edits; candidate edits are merged, deduplicated, ranked, and clipped by a textual learning rate\u2014a per-step edit budget. Every candidate skill must then pass a strict validation gate: it is adopted only if it scores strictly higher than the current skill on the held-out validation split. Rejected edits are not discarded; they enter a rejected-edit buffer that serves as negative feedback for later optimizer calls in the same epoch. On a slower cadence, an epoch-wise slow\/meta update consolidates longer-horizon lessons that single batches cannot reveal (Figure 2). Together, bounded edits, validation gating, and best-version selection keep skill optimization controllable and auditable, so the skill converges instead of drifting.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"586\" height=\"322\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-2.png\" alt=\"Figure 2. The SkillOpt pipeline: trajectory collection, minibatch reflection, bounded text updates, validation gating, and epoch-wise slow\/meta updates jointly constrain skill training.\" class=\"wp-image-1176931\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-2.png 586w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-2-300x165.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt_EXT_FA_Figure-2-240x132.png 240w\" sizes=\"auto, (max-width: 586px) 100vw, 586px\" \/><figcaption class=\"wp-element-caption\">Figure 2. The SkillOpt pipeline: trajectory collection, minibatch reflection, bounded text updates, validation gating, and epoch-wise slow\/meta updates jointly constrain skill training.<\/figcaption><\/figure>\n\n\n\n<h2 id=\"consistent-gains-across-benchmarks-models-and-execution-modes\" class=\"wp-block-heading\">Consistent gains across benchmarks, models, and execution modes<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">We evaluated SkillOpt across six benchmarks (SearchQA, SpreadsheetBench, OfficeQA, DocVQA, LiveMathematicianBench, and ALFWorld), seven target models from frontier-scale GPT-5.5 to the small open-weight Qwen3.5-4B, and three execution modes (direct chat, Codex, and Claude Code). Counting each combination as one evaluation cell, When measured against human-written skills, one-shot LLM skills, Trace2Skill, TextGrad, GEPA, and EvoSkill, SkillOpt delivered the best or tied for -best results on all 52 cells. These performance improvements are unusually large for a method that updates no model weights. With GPT-5.5 in direct chat, SkillOpt raises the six-benchmark average from 58.8 to 82.3, a +23.5-point absolute improvement\u2014and +5.4 points above an oracle that picks the single best competing method per cell. The largest gains appear on procedural benchmarks: SpreadsheetBench rises from 41.8 to 80.7, OfficeQA from 33.1 to 72.1, and LiveMathematicianBench from 37.6 to 66.9. The same interface carries over to agentic loops, lifting GPT-5.5 by +24.8 points inside Codex and +19.1 inside Claude Code over no skill.<\/p>\n\n\n\n\t<div class=\"border-bottom border-top border-gray-300 mt-5 mb-5 msr-promo text-center text-md-left alignwide\" data-bi-aN=\"promo\" data-bi-id=\"670821\">\n\t\t\n\n\t\t<p class=\"msr-promo__label text-gray-800 text-center text-uppercase\">\n\t\t<span class=\"px-4 bg-white display-inline-block font-weight-semibold small\">Spotlight: Microsoft research newsletter<\/span>\n\t<\/p>\n\t\n\t<div class=\"row pt-3 pb-4 align-items-center\">\n\t\t\t\t\t\t<div class=\"msr-promo__media col-12 col-md-5\">\n\t\t\t\t<a class=\"bg-gray-300 display-block\" href=\"https:\/\/info.microsoft.com\/ww-landing-microsoft-research-newsletter.html\" aria-label=\"Microsoft Research Newsletter\" data-bi-cn=\"Microsoft Research Newsletter\" target=\"_blank\">\n\t\t\t\t\t<img decoding=\"async\" class=\"w-100 display-block\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2019\/09\/Newsletter_Banner_08_2019_v1_1920x1080.png\" alt=\"\" \/>\n\t\t\t\t<\/a>\n\t\t\t<\/div>\n\t\t\t\n\t\t\t<div class=\"msr-promo__content p-3 px-5 col-12 col-md\">\n\n\t\t\t\t\t\t\t\t\t<h2 class=\"h4\">Microsoft Research Newsletter<\/h2>\n\t\t\t\t\n\t\t\t\t\t\t\t\t<p id=\"microsoft-research-newsletter\" class=\"large\">Stay connected to the research community at Microsoft.<\/p>\n\t\t\t\t\n\t\t\t\t\t\t\t\t<div class=\"wp-block-buttons justify-content-center justify-content-md-start\">\n\t\t\t\t\t<div class=\"wp-block-button is-style-fill-chevron\">\n\t\t\t\t\t\t<a href=\"https:\/\/info.microsoft.com\/ww-landing-microsoft-research-newsletter.html\" aria-describedby=\"microsoft-research-newsletter\" class=\"btn btn-brand glyph-append glyph-append-chevron-right\" data-bi-cn=\"Microsoft Research Newsletter\" target=\"_blank\">\n\t\t\t\t\t\t\tSubscribe today\t\t\t\t\t\t<\/a>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t<\/div><!--\/.msr-promo__content-->\n\t<\/div><!--\/.msr-promo__inner-wrap-->\n\t<\/div><!--\/.msr-promo-->\n\t\n\n\n<h2 id=\"a-small-model-plus-a-skill-file\" class=\"wp-block-heading\">A small model plus a skill file<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Approaching the next model tier SkillOpt also narrows the gap between small or open-weight models and frontier models\u2014without changing any weights or adding any extra model calls at inference. After optimization, GPT-5.4-mini\u2019s six-benchmark average (64.3) exceeds the no-skill baseline of the larger GPT-5.4 (59.7), and GPT-5.4-nano (57.4) exceeds the no-skill baseline of GPT-5.2 (51.3). Qwen3.5-4B, a 4-billion-parameter open-weight model, surpasses GPT-5.2\u2019s no-skill baseline as well. Gains that once required a larger model can now be approximated by one optimized skill file.<\/p>\n\n\n\n<h2 id=\"skills-that-transfer-train-once-reuse-everywhere\" class=\"wp-block-heading\">Skills that transfer: train once, reuse everywhere<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The optimized skill file captures reusable task-solving procedures rather than instructions overfit to a single model, benchmark, or execution environment. This is why the same skill can still improve performance when transferred across model scales, agent harnesses, and related tasks. In our transfer experiments, skills continued to deliver gains when moved across model scales, across execution harnesses, and to a nearby math benchmark. The clearest example is cross-harness transfer: a spreadsheet skill trained inside Codex, dropped into Claude Code with no further optimization, lifts the no-skill baseline from 22.1 to 81.8 (+59.7)\u2014slightly above the 80.4 achieved by training directly inside Claude Code. Because the two harnesses expose different tool surfaces, this suggests SkillOpt learns general workflow logic, not just harness-specific recipes.<\/p>\n\n\n\n<h2 id=\"compact-readable-and-built-from-very-few-accepted-edits\" class=\"wp-block-heading\">Compact, readable, and built from very few accepted edits<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The deployed artifact, best_ skill.md , is neither an opaque parameter blob nor an ever-growing log. Across six case studies, the median final skill length is roughly 920 tokens, and because the validation gate rejects most proposals, only one to four edits are accepted into the final file. OfficeQA\u2019s +39.0-point gain comes from a single accepted edit. The learned rules read like a seasoned practitioner\u2019s advice. Component ablations confirm that the controls do the work: removing the rejected-edit buffer lowers scores on all three ablation benchmarks, and removing both the meta skill and the slow update drops SpreadsheetBench from 77.5 to 55.0. A new adaptation layer for the agent era SkillOpt points to a lighter-weight path for domain-adapting agents: instead of fine-tuning weights, hard-coding task logic, or hand-tuning prompts, teams can train a small, versionable, auditable natural-language skill layer\u2014wherever automatic evaluation or a reliable verifier exists.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">By bringing learning rates, schedules, validation splits, rejected samples, and slow updates to agent skills, SkillOpt suggests that training need not be limited to model weights.&nbsp;Procedural knowledge outside the model can also be optimized.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">When that process is controlled, validated, and recorded, a natural-language skill becomes a stable, transferable, and reversible adapter between frontier-model capability and real-world workloads. Read the full paper, visit the project page at&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/aka.ms\/skillopt\">aka.ms\/skillopt<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, or explore the SkillOpt GitHub repository at&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/microsoft\/SkillOpt\" target=\"_blank\" rel=\"noopener noreferrer\">github.com\/microsoft\/SkillOpt<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>. Teams building agentic workflows can use SkillOpt as a foundation for training reusable skills against their own tasks and verifiers. See also our companion project, SkillLens.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-fe48e5de wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/skillopt-executive-strategy-for-self-evolving-agent-skills\/\">Paper<\/a><\/div>\n\n\n\n<div class=\"wp-block-button is-style-fill-github\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/github.com\/microsoft\/SkillOpt\">GitHub<\/a><\/div>\n\n\n\n<div class=\"wp-block-button is-style-fill\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/microsoft.github.io\/SkillLens\/\">SkillLens Project Page<\/a><\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>AI agents often fail because their instructions, or skills, are manually modified with no guarantee of improvement. Learn how SkillOpt turns skill editing into a training process, making agent behavior more reliable without changing model weights.<\/p>\n","protected":false},"author":43868,"featured_media":1176928,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":[{"type":"user_nicename","value":"Yifan Yang","user_id":"41539"},{"type":"guest","value":"xuemei-gao","user_id":"1177062"},{"type":"user_nicename","value":"Qi Dai","user_id":"36689"},{"type":"user_nicename","value":"Bei Liu","user_id":"38889"},{"type":"user_nicename","value":"Kai Qiu","user_id":"38988"},{"type":"user_nicename","value":"Dongdong Chen","user_id":"40198"},{"type":"user_nicename","value":"Chong Luo","user_id":"31450"}],"msr_hide_image_in_river":0,"footnotes":""},"categories":[1],"tags":[],"research-area":[13556,13562],"msr-region":[],"msr-event-type":[],"msr-locale":[268875],"msr-post-option":[243984],"msr-impact-theme":[],"msr-promo-type":[],"msr-podcast-series":[],"class_list":["post-1176927","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-research-blog","msr-research-area-artificial-intelligence","msr-research-area-computer-vision","msr-locale-en_us","msr-post-option-blog-homepage-featured"],"msr_event_details":{"start":"","end":"","location":""},"podcast_url":"","podcast_episode":"","msr_research_lab":[199560,1012650],"msr_impact_theme":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-academic-programs":[],"related-groups":[],"related-projects":[],"related-events":[],"related-researchers":[{"type":"user_nicename","value":"Yifan Yang","user_id":41539,"display_name":"Yifan Yang","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/yifanyang\/\" aria-label=\"Visit the profile page for Yifan Yang\">Yifan Yang<\/a>","is_active":false,"last_first":"Yang, Yifan","people_section":0,"alias":"yifanyang"},{"type":"guest","value":"xuemei-gao","user_id":1177062,"display_name":"Xuemei Gao","author_link":"Xuemei Gao","is_active":true,"last_first":"Gao, Xuemei","people_section":0,"alias":"xuemei-gao"},{"type":"user_nicename","value":"Qi Dai","user_id":36689,"display_name":"Qi Dai","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/qid\/\" aria-label=\"Visit the profile page for Qi Dai\">Qi Dai<\/a>","is_active":false,"last_first":"Dai, Qi","people_section":0,"alias":"qid"},{"type":"user_nicename","value":"Bei Liu","user_id":38889,"display_name":"Bei Liu","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/libei\/\" aria-label=\"Visit the profile page for Bei Liu\">Bei Liu<\/a>","is_active":false,"last_first":"Liu, Bei","people_section":0,"alias":"libei"},{"type":"user_nicename","value":"Kai Qiu","user_id":38988,"display_name":"Kai Qiu","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/kaqiu\/\" aria-label=\"Visit the profile page for Kai Qiu\">Kai Qiu<\/a>","is_active":false,"last_first":"Qiu, Kai","people_section":0,"alias":"kaqiu"},{"type":"user_nicename","value":"Dongdong Chen","user_id":40198,"display_name":"Dongdong Chen","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/dochen\/\" aria-label=\"Visit the profile page for Dongdong Chen\">Dongdong Chen<\/a>","is_active":false,"last_first":"Chen, Dongdong","people_section":0,"alias":"dochen"},{"type":"user_nicename","value":"Chong Luo","user_id":31450,"display_name":"Chong Luo","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/cluo\/\" aria-label=\"Visit the profile page for Chong Luo\">Chong Luo<\/a>","is_active":false,"last_first":"Luo, Chong","people_section":0,"alias":"cluo"}],"msr_type":"Post","featured_image_thumbnail":"<img width=\"960\" height=\"540\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-960x540.jpg\" class=\"img-object-cover\" alt=\"SkillOpt blog | three white line icons on an abstract green background | shield icon, gear icon, circle with checkmark icon\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1536x865.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-2048x1153.jpg 2048w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2026\/06\/SkillOpt-BlogHeroFeature-1400x788-1-1920x1080.jpg 1920w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/>","byline":"","formattedDate":"June 30, 2026","formattedExcerpt":"AI agents often fail because their instructions, or skills, are manually modified with no guarantee of improvement. Learn how SkillOpt turns skill editing into a training process, making agent behavior more reliable without changing model weights.","locale":{"slug":"en_us","name":"English","native":"","english":"English"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1176927","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/43868"}],"replies":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/comments?post=1176927"}],"version-history":[{"count":25,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1176927\/revisions"}],"predecessor-version":[{"id":1177177,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1176927\/revisions\/1177177"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1176928"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1176927"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/categories?post=1176927"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/tags?post=1176927"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1176927"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=1176927"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=1176927"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1176927"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1176927"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1176927"},{"taxonomy":"msr-promo-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-promo-type?post=1176927"},{"taxonomy":"msr-podcast-series","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-podcast-series?post=1176927"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}