{"id":1190639,"date":"2026-10-21T10:26:02","date_gmt":"2026-10-21T17:26:02","guid":{"rendered":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/publication\/coding-agents-are-strong-prompt-optimizers\/"},"modified":"2026-10-22T15:32:28","modified_gmt":"2026-10-22T22:32:28","slug":"coding-agents-are-strong-prompt-optimizers","status":"publish","type":"msr-research-item","link":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/publication\/coding-agents-are-strong-prompt-optimizers\/","title":{"rendered":"Coding Agents are Strong Prompt Optimizers"},"content":{"rendered":"\n\n\n<p class=\"wp-block-paragraph\">Search-based prompt optimizers improve prompts through iterative search: they propose edits, execute fresh rollouts, score the resulting trajectories, and retain only edits that improve a validation metric. We show that this optimization loop is unnecessary. Given only a static corpus of agent trajectories, an off-the-shelf coding agent can directly synthesize an optimized prompt, requiring neither environment access nor validation data. We call this approach textit{Coding-Agent Skill Distillation} (CASD). The key insight is reflection scope. Rather than reasoning over a small batch of trajectories at each optimization step, the coding agent writes and executes analysis code to compute corpus-wide statistics, identifies systematic failure modes, inspects representative episodes, and distills the resulting insights into behavioral rules. Across four agentic benchmarks (ALFWorld, <math><mrow><msup><mi>\u03c4<\/mi><mn>2<\/mn><\/msup><\/mrow><\/math>-bench retail and telecom, and SpreadsheetBench-Verified), under matched data access, a single CASD pass outperforms GEPA, a state-of-the-art reflective prompt optimizer, on three of four benchmarks and outperforms validation-gated reflective search (SkillOpt) on all four, improving the unoptimized baseline by 16.6 percentage points on average versus 10.9 for GEPA and 5.3 for SkillOpt. Because CASD performs a single offline analysis pass rather than iterative search, producing an optimized prompt costs approximately $1.60&#8212;over $22times$ cheaper than validation-gated search. Even when competing methods are granted additional validation data and unrestricted environment access, CASD remains ahead on two of four benchmarks. These results suggest that corpus-scale statistical reflection is a viable alternative to iterative search for prompt optimization.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Search-based prompt optimizers improve prompts through iterative search: they propose edits, execute fresh rollouts, score the resulting trajectories, and retain only edits that improve a validation metric. We show that this optimization loop is unnecessary. Given only a static corpus of agent trajectories, an off-the-shelf coding agent can directly synthesize an optimized prompt, requiring neither [&hellip;]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":[{"type":"text","value":"Agamdeep Singh","user_id":0},{"type":"user_nicename","value":"Srishti Gautam","user_id":"43800"},{"type":"user_nicename","value":"Priyanshu Gupta","user_id":"42237"},{"type":"text","value":"N. Mehrotra","user_id":0},{"type":"text","value":"Tanmay Bakshi","user_id":0},{"type":"user_nicename","value":"Sumit Gulwani","user_id":"33755"}],"msr_publishername":"","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"","msr_editors":"","msr_how_published":"arXiv","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"","msr_doi":"","msr_arxiv_id":"2609.26261","msr_mag_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_release_tracker_id":"","msr_highlight_type":"","msr_date_display_format":"","msr_main_download_label":"","msr_external_link_label":"","msr_doi_label":"","msr_published_date":"2026-08-13","msr_startdate":"","msr_presentation_date":"","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_year":2026,"msr_month":8,"msr_day":13,"msr_microsoftintellectualproperty":false,"msr_pub_id":"a9ba521e28745b38e5bc3f8dcaf67929dd01bb78","msr_publication_uploader":[{"type":"url","title":"https:\/\/arxiv.org\/abs\/2609.26261","label_id":252679,"id":false,"viewUrl":false}],"msr_related_uploader":[],"msr_original_fields_of_study":[],"msr_s2_paper_id":"a9ba521e28745b38e5bc3f8dcaf67929dd01bb78","msr_s2_pdf_url":"","msr_citation_count_updated":"","msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":40,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[{"provider":"s2","id":"a9ba521e28745b38e5bc3f8dcaf67929dd01bb78"},{"provider":"arxiv","id":"2609.26261"},{"provider":"corpusid","id":"292227444"}],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[],"research-area":[13556,13545,13555],"msr-publication-type":[270373],"msr-publisher":[],"msr-publication-cta":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[256039,246691],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1190639","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-human-language-technologies","msr-research-area-search-information-retrieval","msr-locale-en_us","msr-field-of-study-artifcial-intelligence","msr-field-of-study-computer-science"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2026-08-13","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"arXiv","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":0,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"url","title":"https:\/\/arxiv.org\/abs\/2609.26261","label_id":252679,"id":false,"viewUrl":false}],"msr_related_uploader":[],"msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"a9ba521e28745b38e5bc3f8dcaf67929dd01bb78","msr_influential_citations":0,"msr_reference_count":40,"msr_arxiv_id":"2609.26261","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[],"msr-author-ordering":[{"type":"text","value":"Agamdeep Singh","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Srishti Gautam","user_id":43800,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Srishti Gautam"},{"type":"user_nicename","value":"Priyanshu Gupta","user_id":42237,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Priyanshu Gupta"},{"type":"text","value":"N. Mehrotra","user_id":0,"rest_url":false},{"type":"text","value":"Tanmay Bakshi","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Sumit Gulwani","user_id":33755,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Sumit Gulwani"}],"msr_impact_theme":[],"msr_research_lab":[199562],"msr_event":[],"msr_group":[],"msr_project":[],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"misc","related_content":[],"_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1190639","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":3,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1190639\/revisions"}],"predecessor-version":[{"id":1190854,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1190639\/revisions\/1190854"}],"wp:attachment":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1190639"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=1190639"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1190639"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=1190639"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=1190639"},{"taxonomy":"msr-publication-cta","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-cta?post=1190639"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=1190639"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1190639"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1190639"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=1190639"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=1190639"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=1190639"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1190639"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1190639"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}