repomap-auto-on
Enable automatic repo map updates. When enabled, REPOMAP.md is incrementally updated every time a file is edited.
Find an Agent Skill for your task. Read the guide and copy the install command.
Updated 11-17 00:07
41,460 results
Enable automatic repo map updates. When enabled, REPOMAP.md is incrementally updated every time a file is edited.
Pull a PR from GitHub, run parallel review agents (code reuse, quality, efficiency, spec adherence), apply local fixes, and write a self-contained HTML review…
Sketch types, signatures, and module structure before code, then stay in the loop while implementation fills in. Use for /architect, 'architect this', 'design…
Spawn the comment-sicko subagent, fix accepted findings, and offer encodings for claimed constraints.
Run an extremely strict maintainability review for abstraction quality, giant files, and spaghetti-condition growth. Use for a thermo-nuclear code quality…
TypeScript best practices. Use when reading or editing any .ts or .tsx file.
Enforces high-variation, low-card, non-generic frontend design output for coding agents.
Execute a specific implementation phase from a plan in docs/plans, run corresponding validations, and update that phase verification checkboxes as tasks…
Consolidated project stack skill with integration patterns — code-mode (analyzes manifests) or compose-mode (synthesizes from existing skills + architecture…
Validate AI-generated PR review comments — check whether they are accurate or hallucinated by cross-referencing each comment against the actual code changes.…
Matmul/Cube/GEMM/BMM single-operator direct invocation generation (Blaze/tensor_api path). This skill must be used when the user mentions Blaze. Covers selection, modification, Tiling, and troubleshooting of the three templates: pure AIC / StreamK / FixpOpti. Not applicable to Vector element-wise/reduction operators or non-Blaze paths.
A performance optimization expert skilled at writing efficient Triton operators on the Ascend NPU platform. Optimize Triton code step by step in strict order, trying only one optimization point at a time, ensuring functional consistency and precision consistency before and after optimization. ⚠️ Only the optimization methods specified by this skill may be used; any optimization method beyond this skill is prohibited. Trigger: when the user needs to optimize Ascend...
Triton Ascend Operator Algorithm Sketch Design Skill — designs a high-quality algorithm sketch based on the task description, used to guide subsequent code generation. Supports both first-time design and iterative optimization based on historical context. Trigger: used when the user needs to design an algorithm sketch for a Triton Ascend operator or iterate on an existing sketch.
Operator Code Verification Skill — Verify generated kernel code according to the standard verification workflow. Create the verification project files, call scripts/verify.py to run verification, and after verification passes, call scripts/benchmark.py to perform performance testing and collect the results. Trigger: used when the user needs to verify the functional correctness of Triton operator code or collect its performance data.
Extract operator implementations from the user's PyTorch/Python code and build them into standardized task files in the operator task format. Supports two modes: single case (a single self-contained .py, get_inputs returns a single group) and multi case (.py paired with a .json of the same name, get_input_groups returns multiple groups). Trigger: when the user needs to convert…
Provides step-by-step code migration procedures from GPU/CPU to Ascend NPU. Invoke when performing actual code adaptation, interface replacement, or debugging…
Provides torch\_npu API compatibility reference and interface mapping knowledge. Invoke when checking NPU interface support, finding equivalent replacements,…
Perform a Roofline check of compute bottlenecks on the Ascend profiler's kernel_details.csv: filter operators with aic_mac_ratio≥80%, compare theoretical time with actual time, determine whether optimization is needed, and give investigation directions. When the user mentions…
Write README-style technical documentation for AscendC custom operators by reading local source files and adapting an existing template. Use when Codex needs…
Systematic and safe code refactoring skill. Able to identify code smells, provide refactoring techniques, support creating plans, executing refactorings, and verifying results. Suitable for scenarios such as code quality improvement, technical debt cleanup, maintainability improvement, code optimization, and code improvement. Triggered when the user requests refactoring code, optimizing code quality, fixing code issues, or cleaning up technical debt.
Analyzes user-provided error messages, logs, and environment information to identify root causes and generate customer-friendly responses for Ascend NPU…
Fetch and query data from GitCode platform via its REST API: repositories, branches, issues, pull requests, commits, tags, users, organizations, search,…
Generate release notes for GitCode repositories from commits (by tag range or since-date), grouped as feat/fix/docs/other, output Markdown for Release pages. By...
Model resource collection. Search online for the latest free/discounted AI model resource information (such as major cloud providers' promotional activities, free APIs, coding plans, etc.), organize it, and generate a configuration guide document. Applicable to scenarios where users ask "what cheap models do you recommend", "what free models are there", "model discounts", etc.