Large Language Models (LLMs) increasingly act as agents whose procedural knowledge is stored in reusable skill packages and loaded at inference time. As skill libraries grow, a central challenge is to expose the smallest sufficient executable context under a limited context budget. Existing systems struggle to reuse routines below the whole-skill level, preserve procedural contracts during compression, keep compressed routines executable and expandable, and update the compressed library as skills evolve. These challenges reveal a unit mismatch: skills are retrieved as packages, compressed as text, and converted into execution graphs only after retrieval, whereas reliable reuse requires a contract-bearing procedural unit. We propose SkillZip, an execution-aware procedural abstraction framework that performs contract-preserving compression over section-level graphs. SkillZip rewrites recurring contract-valid motifs into reversible ported macros while preserving boundary signatures, dependency closure, verifier reachability, and source-level expansion. At inference time, it hydrates a compact, dependency-closed context and expands macros only when required. ReZip further integrates new skills and revises risky macros using execution evidence. </p>\n","updatedAt":"2026-08-07T10:08:29.035Z","author":{"_id":"664455689d51633c52e39308","avatarUrl":"/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg","fullname":"Tan","name":"stevettt","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8877000212669373},"editors":["stevettt"],"editorAvatarUrls":["/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg"],"reactions":[],"isReport":false}},{"id":"6a7daf9df05652c3620e2946","author":{"_id":"664455689d51633c52e39308","avatarUrl":"/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg","fullname":"Tan","name":"stevettt","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2026-08-13T11:50:53.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"SkillZip compresses reusable procedural skills into contract-preserving, executable graph units to enable efficient retrieval and expansion under limited context budgets.","html":"<p>SkillZip compresses reusable procedural skills into contract-preserving, executable graph units to enable efficient retrieval and expansion under limited context budgets.</p>\n","updatedAt":"2026-08-13T11:50:53.307Z","author":{"_id":"664455689d51633c52e39308","avatarUrl":"/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg","fullname":"Tan","name":"stevettt","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8293377757072449},"editors":["stevettt"],"editorAvatarUrls":["/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg"],"reactions":[],"isReport":false}},{"id":"6a7dc74f737d73e6365ff819","author":{"_id":"6a69eb40da85bffbe935dc52","avatarUrl":"/avatars/f7af2d579a31754981d459c35b9495c4.svg","fullname":"Mary Hernandez","name":"hernandez-m","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2026-08-13T13:31:59.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Really nice work! SkillZip is a simple and practical idea for making large skill libraries more efficient. The section-level design is especially interesting.","html":"<p>Really nice work! SkillZip is a simple and practical idea for making large skill libraries more efficient. The section-level design is especially interesting.</p>\n","updatedAt":"2026-08-13T13:31:59.977Z","author":{"_id":"6a69eb40da85bffbe935dc52","avatarUrl":"/avatars/f7af2d579a31754981d459c35b9495c4.svg","fullname":"Mary Hernandez","name":"hernandez-m","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9107043147087097},"editors":["hernandez-m"],"editorAvatarUrls":["/avatars/f7af2d579a31754981d459c35b9495c4.svg"],"reactions":[{"reaction":"❤️","users":["stevettt"],"count":1}],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.05604","authors":[{"_id":"6a755e0ee1228e04b32382c1","user":{"_id":"664455689d51633c52e39308","avatarUrl":"/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg","isPro":false,"fullname":"Tan","user":"stevettt","type":"user","name":"stevettt"},"name":"Xingyu Tan","status":"claimed_verified","statusLastChangedAt":"2026-08-13T08:45:04.950Z","hidden":true},{"_id":"6a755e0ee1228e04b32382c2","name":"Xiaoyang Wang","hidden":false},{"_id":"6a755e0ee1228e04b32382c3","name":"Qing Liu","hidden":false},{"_id":"6a755e0ee1228e04b32382c4","name":"Xiwei Xu","hidden":false},{"_id":"6a755e0ee1228e04b32382c5","name":"Xin Yuan","hidden":false},{"_id":"6a755e0ee1228e04b32382c6","name":"Liming Zhu","hidden":false},{"_id":"6a755e0ee1228e04b32382c7","name":"Wenjie Zhang","hidden":false}],"publishedAt":"2026-08-06T00:00:00.000Z","submittedOnDailyAt":"2026-08-13T00:00:00.000Z","title":"SkillZip: Contract-Preserving Graph Compression for Scalable Agent Skill Libraries","submittedOnDailyBy":{"_id":"664455689d51633c52e39308","avatarUrl":"/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg","isPro":false,"fullname":"Tan","user":"stevettt","type":"user","name":"stevettt"},"summary":"Large Language Models (LLMs) increasingly act as agents whose procedural knowledge is stored in reusable skill packages and loaded at inference time. As skill libraries grow, a central challenge is to expose the smallest sufficient executable context under a limited context budget. Existing systems struggle to reuse routines below the whole-skill level, preserve procedural contracts during compression, keep compressed routines executable and expandable, and update the compressed library as skills evolve. These challenges reveal a unit mismatch: skills are retrieved as packages, compressed as text, and converted into execution graphs only after retrieval, whereas reliable reuse requires a contract-bearing procedural unit. We propose SkillZip, an execution-aware procedural abstraction framework that performs contract-preserving compression over section-level graphs. SkillZip rewrites recurring contract-valid motifs into reversible ported macros while preserving boundary signatures, dependency closure, verifier reachability, and source-level expansion. At inference time, it hydrates a compact, dependency-closed context and expands macros only when required. ReZip further integrates new skills and revises risky macros using execution evidence. Comprehensive experiments1 on technical and embodied agent benchmarks show SkillZip consistently outperforms the strongest baseline by up to 12.2 points, while achieving a 3.46x compression ratio with 99.2% dependency preservation and 98.7% verifier reachability. Scaling analyses further confirm robust retrieval across skill libraries ranging from 200 to 100K skills.","upvotes":29,"discussionId":"6a755e0ee1228e04b32382c8","ai_summary":"SkillZip compresses reusable procedural skills into contract-preserving, executable graph units to enable efficient retrieval and expansion under limited context budgets.","ai_keywords":["Large Language Models","skill packages","procedural abstraction","contract-preserving compression","section-level graphs","ported macros","dependency closure","verifier reachability","SkillZip","ReZip"],"ai_summary_model":"thinkingmachines/Inkling-Small"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"664455689d51633c52e39308","avatarUrl":"/avatars/74bd358b0d1d5c7f37ed3500ac446cd8.svg","isPro":false,"fullname":"Tan","user":"stevettt","type":"user"},{"_id":"6968b9286ed97e3f8250a9c9","avatarUrl":"/avatars/4e500acf929e40da58d8856a5f2a1e66.svg","isPro":false,"fullname":"XJ Z","user":"zaixjun","type":"user"},{"_id":"6a5ecbe7cede9269bc2de600","avatarUrl":"/avatars/7d21689dc435cf2df4889882a8934932.svg","isPro":false,"fullname":"Peiting Xie","user":"Peitingx","type":"user"},{"_id":"643d746bd455a0d0a0a881b3","avatarUrl":"/avatars/f103799558db803a1810b3dc5eb31c87.svg","isPro":false,"fullname":"Chengkai Huang","user":"Chengkai","type":"user"},{"_id":"6a7db384cbebb41caacd56f4","avatarUrl":"/avatars/80717ab21967b1bed000063d23762ba6.svg","isPro":false,"fullname":"haozhe yin","user":"dypoli","type":"user"},{"_id":"6a7db99af36a729cd188bf59","avatarUrl":"/avatars/7f25c45685b0ce1d0ab7194972a898ea.svg","isPro":false,"fullname":"Kaiyu Chen","user":"kaiyuchen1","type":"user"},{"_id":"6833bcc07e0c637c71d00a03","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/SqnuYnxNOmyINfvIl75WP.png","isPro":false,"fullname":"Zheng","user":"Liangwei1037","type":"user"},{"_id":"659bcee4654fe4eb0a560fdf","avatarUrl":"/avatars/7c0417bfb37380521639070f53c5eb25.svg","isPro":false,"fullname":"Fan Li","user":"Philley","type":"user"},{"_id":"6a69eb40da85bffbe935dc52","avatarUrl":"/avatars/f7af2d579a31754981d459c35b9495c4.svg","isPro":false,"fullname":"Mary Hernandez","user":"hernandez-m","type":"user"},{"_id":"6a69eb62c5e36f5d1b3c77e4","avatarUrl":"/avatars/8481802389ca8628ac440d9b926ca9ac.svg","isPro":false,"fullname":"William Lopez","user":"wlopez-nlp","type":"user"},{"_id":"6a69eb859ef85244dc4ca374","avatarUrl":"/avatars/21fc260ba33de7f03d663e84fba9199b.svg","isPro":false,"fullname":"Anthony Sanchez","user":"anthony-code","type":"user"},{"_id":"6a69ec3496892fc204f5f088","avatarUrl":"/avatars/2a4ce4c0bce3a56a7e3eea365b0c7291.svg","isPro":false,"fullname":"Barbara Lopez","user":"barbara-ai","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.05604.md","query":{}}">
SkillZip: Contract-Preserving Graph Compression for Scalable Agent Skill Libraries
Published on Aug 6
· Submitted by Tan on Aug 13 Abstract
SkillZip compresses reusable procedural skills into contract-preserving, executable graph units to enable efficient retrieval and expansion under limited context budgets.
Large Language Models (LLMs) increasingly act as agents whose procedural knowledge is stored in reusable skill packages and loaded at inference time. As skill libraries grow, a central challenge is to expose the smallest sufficient executable context under a limited context budget. Existing systems struggle to reuse routines below the whole-skill level, preserve procedural contracts during compression, keep compressed routines executable and expandable, and update the compressed library as skills evolve. These challenges reveal a unit mismatch: skills are retrieved as packages, compressed as text, and converted into execution graphs only after retrieval, whereas reliable reuse requires a contract-bearing procedural unit. We propose SkillZip, an execution-aware procedural abstraction framework that performs contract-preserving compression over section-level graphs. SkillZip rewrites recurring contract-valid motifs into reversible ported macros while preserving boundary signatures, dependency closure, verifier reachability, and source-level expansion. At inference time, it hydrates a compact, dependency-closed context and expands macros only when required. ReZip further integrates new skills and revises risky macros using execution evidence. Comprehensive experiments1 on technical and embodied agent benchmarks show SkillZip consistently outperforms the strongest baseline by up to 12.2 points, while achieving a 3.46x compression ratio with 99.2% dependency preservation and 98.7% verifier reachability. Scaling analyses further confirm robust retrieval across skill libraries ranging from 200 to 100K skills.
Community
Large Language Models (LLMs) increasingly act as agents whose procedural knowledge is stored in reusable skill packages and loaded at inference time. As skill libraries grow, a central challenge is to expose the smallest sufficient executable context under a limited context budget. Existing systems struggle to reuse routines below the whole-skill level, preserve procedural contracts during compression, keep compressed routines executable and expandable, and update the compressed library as skills evolve. These challenges reveal a unit mismatch: skills are retrieved as packages, compressed as text, and converted into execution graphs only after retrieval, whereas reliable reuse requires a contract-bearing procedural unit. We propose SkillZip, an execution-aware procedural abstraction framework that performs contract-preserving compression over section-level graphs. SkillZip rewrites recurring contract-valid motifs into reversible ported macros while preserving boundary signatures, dependency closure, verifier reachability, and source-level expansion. At inference time, it hydrates a compact, dependency-closed context and expands macros only when required. ReZip further integrates new skills and revises risky macros using execution evidence.
SkillZip compresses reusable procedural skills into contract-preserving, executable graph units to enable efficient retrieval and expansion under limited context budgets.
Really nice work! SkillZip is a simple and practical idea for making large skill libraries more efficient. The section-level design is especially interesting.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2608.05604 in a model README.md to link it from this page.
Cite arxiv.org/abs/2608.05604 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2608.05604 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.