Explore publicações revisadas no X, quem as compartilha e os projetos mencionados.
A contagem inclui apenas publicações sobre Skills coletadas e revisadas. Um dia vazio indica que nada foi coletado aqui. Datas em UTC.
208 publicações coletadas
Halk shares orchestrate-work in response to a request for agent-to-agent handoffs. The Skill defines a goal, routes work across agent roles, and requires evidence before accepting the result.
Qiaomu shares qiaomu-codex-imagegen, a Skill with MCP and CLI routes for calling Codex's built-in image generation from another agent.
The author shares materials-simulation-skills after its inclusion in Kimi’s Claude Code collection. The repository gives agents instructions for numerical stability, convergence and simulation workflows, with the stated aim of getting the simulation right rather than merely running the code.
The creator introduces biaodian, a Chinese punctuation Skill grounded in GB/T 15834-2011. A mechanical scan and semantic review produce an issue list, explanations with source references, and corrected text. It changes punctuation without rewriting words, numbers or style.
The author shares douyin-download for fetching one Douyin work from share text, a link or an ID. It can save the available original images, clean video stream, Live Photo media and BGM, with metadata and SHA256 checks. It requires a logged-in browser and does not collect creator profiles in bulk.
Chooses writing, diagrams, interactive pages, or video around a learning goal and audience, then gives a reusable prompt with its rationale. Actual artifact creation is a separate request.
This recommendation highlights a Codex Skill that turns keywords, articles, files, and screenshots into Xiaohongshu knowledge posts. It researches the topic before planning the pages and creating the image cards.
The gh-stack CLI extension includes an Agent Skill for splitting dependent changes into a stack of branches and pull requests. Each PR uses the branch below it as its base, so reviewers can examine one layer at a time.
Antoine shares three updates to SwiftUI Expert Skill for adapting iPhone Duo layouts across fold states. The linked 5.2.1 release extends guidance drawn from Apple resources, Xcode’s bundled Skill and his own app work. The Skill is available under MIT.
Adds tester-army/e2e to selected projects, assigning one coding subagent per project and a verifier for each loop to cover flows, run tests, and report failures.
A skill for tuning self-hosted LLM inference around the model, hardware, API requirements, and serving workload, including parallelism, KV tiers, and prefill/decode separation.
Packages a character-replacement video workflow, from reference-video checks and reusable character profiles to prompts and comparison clips.
A recommendation of yomiyasu for rewriting stiff AI-generated Japanese. The Skill works on technical articles, business documents, and essays while preserving the source’s meaning and constraints.
Turns a photo of six-line guitar tablature with jianpu into a play-along page, with a moving cursor, synchronized melody and accompaniment, tempo control, and loops.
This share highlights qisi-video-remix, which adapts a reference video into a new script, character and scene plans, storyboards, and copyable prompts for each video segment. The repository’s default deliverable is a production plan.
Claritymaxx's explain skill reads the source material and builds a mental model before choosing prose, a diagram, or an HTML explainer. It separates what the source shows from what the explanation infers.
A recommendation of book-to-skill, which turns technical books and document collections into a reusable Agent Skill with frameworks and chapter references that an agent can load on demand.
A recommendation of ip-as-logo for rounded square mascot images with simple silhouettes and a solid background. The workflow proposes design directions before generating a set of independent candidates.
Anime Reaction GIF makes looping anime reaction stickers from generated character poses and timed holds. The creator shares a catgirl sneeze example and the open-source Codex Skill.
Josh Holtz shares respond-clearly for direct, scannable agent replies with clear next steps.