--- title: "Claim Ledger: Skills Are Package Management for Your AI" description: "There are millions of agent skill files to choose from. The hard part is deciding which deserve trust, how to version them, and how to retire them." author: "Harry Floyd" publication: "The Durability Curve" canonical: "https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/" essay: "https://durabilitycurve.com/blog/skills-are-package-management-for-your-ai/" substack: "https://harryfloyd.substack.com/p/skills-are-package-management" published: "2026-06-07" ledger_date: "2026-10-02" law: "Law III" claims: 8 removed_before_publication: 0 --- # Claim Ledger: Skills Are Package Management for Your AI *There are millions of agent skill files to choose from. The hard part is deciding which deserve trust, how to version them, and how to retire them.* Essay: https://durabilitycurve.com/blog/skills-are-package-management-for-your-ai/ Ledger (canonical, cite this): https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/ Published 2026-06-07 · last verified 2026-10-02 ## The evidence, row by row Labels: VERIFIED = checked against the primary source itself by a checker who did not draft the piece, usually with the exact words, where they sit and the date read · EXECUTED = a run or observation done for the piece (code, a query, a count, or a check made live in an app); the claim is what it returned; run records are not published · CHECKED = checked against its source by a separate checker in the older ledger format, which recorded no quote, locator or date · REPORTED = not verified word for word against a primary source (secondary, not openable in full, or supported only in part); the essay words it accordingly · STRUCK = drafted, checked and removed before publication · EXCLUDED = considered and deliberately left out. ### S1. SkillsMP's live index contained 3,337,075 collected public SKILL.md files on 2 October 2026 - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S1 - Quote: 3,337,075 Collected SKILL.md files - Source: SkillsMP, Agent Skills Marketplace, homepage hero count - URL: https://skillsmp.com/ ### S2. Anthropic published Agent Skills as an open standard intended for cross-platform portability - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S2 - Quote: We've published Agent Skills as an open standard for cross-platform portability. - Source: Anthropic, Equipping agents for the real world with Agent Skills, 18 December 2025 update - URL: https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills ### S3. The Agent Skills format uses SKILL.md instructions that an agent loads when a task matches the skill description - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S3 - Quote: When a task matches a skill’s description, the agent reads the full SKILL.md instructions into context. - Source: Agent Skills, Overview, How do Agent Skills work? — Activation - URL: https://agentskills.io/home ### S4. Agent Skills are supported across a broad set of AI tools and agentic clients - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S4 - Quote: Agent Skills are supported by a large number of AI tools and agentic clients - Source: Agent Skills, Overview and official Client Showcase registry, Where can I use Agent Skills?; linked client registry counted 46 entries on 2026-10-02 - URL: https://agentskills.io/home ### S5. Recent security research frames third-party agent skills as a software-supply-chain risk because developers may install and execute them with implicit trust - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S5 - Quote: developers install and execute third-party skills with implicit trust and no formal analysis - Source: Bhardwaj, Formal Analysis and Supply Chain Security for Agentic AI Skills, arXiv:2603.00195v2, §1 Introduction, lines 205–206 - URL: https://arxiv.org/html/2603.00195v2 ### S6. The same paper defines an Agent Skill Bill of Materials using CycloneDX and records identity/version/hash, capabilities, dependencies, trust metadata and vulnerability status - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S6 - Quote: An ASBOM is a CycloneDX 1.6-compatible document that records - Source: Bhardwaj, Formal Analysis and Supply Chain Security for Agentic AI Skills, arXiv:2603.00195v2, §6.6, Definition 6.8, lines 965–989 - URL: https://arxiv.org/html/2603.00195v2 ### S7. SkillsBench's current release reports that curated Skills raise mean resolution rate from 33.9% to 50.5%, a 16.6-point lift, across the 18-configuration paper aggregate - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S7 - Quote: In the 18-configuration paper aggregate … The mean resolution rate rises from 33.9% to 50.5% (+16.6 points - Source: SkillsBench 1.1: Agent Skills Benchmark Release, Results - URL: https://www.skillsbench.ai/blogs/skillsbench-1-1 ### S8. On SkillsBench's three dedicated self-generated-Skills configurations, self-generated Skills underperformed the no-Skills baseline by 8.1 to 11.5 points - Status: VERIFIED (read 2026-10-02) - Row link: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-S8 - Quote: Self-generated Skills change resolution rate by −8.1 points for Claude Code + Opus 4.7, −11.3 points for Codex + GPT-5.5, and −11.5 points - Source: SkillsBench 1.1: Agent Skills Benchmark Release, Self-Generated Vs. Curated Skills - URL: https://www.skillsbench.ai/blogs/skillsbench-1-1 ## Cite - A claim: name the original source (and its locator) first, then the row it was checked in, e.g. ", . Checked in: Claim Ledger, "Skills Are Package Management for Your AI", The Durability Curve, row , https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/#c-." - The essay: Floyd, Harry (2026). Skills Are Package Management for Your AI. The Durability Curve. https://durabilitycurve.com/blog/skills-are-package-management-for-your-ai/ - This ledger: Floyd, Harry (2026). Claim Ledger: Skills Are Package Management for Your AI. The Durability Curve. https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai/ - Downloads: https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai.csv · https://durabilitycurve.com/claims/skills-are-package-management-for-your-ai.json Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Reuse it, quote it or train on it, with credit to The Durability Curve and a link. Open to AI (robots.txt: search=yes, ai-input=yes, ai-train=yes).