Loading...
Loading...
Found 516 Skills
Safe experimentation framework for AI agents. Creates isolated sandbox environments for trying new features, testing approaches, and exploring solutions without polluting the main codebase. USE WHEN: Agent needs to try something uncertain, explore multiple approaches, test a new library, prototype a feature, or run a technical spike before committing to implementation. PRIMARY TRIGGERS: "experiment with" = Setup sandbox + run experiment "try this approach" = Quick experiment in sandbox "spike" / "POC" / "prototype" = Time-boxed technical investigation "tinker" / "tinkering mode" = Enter experimentation workflow "explore options" = Multi-approach comparison in sandbox NOT FOR: Debugging (use debugger), testing (use test runner), or committed feature work (use git branches). DIFFERENTIATOR: Unlike git branches (for committed direction), tinkering is for "I don't know if this will work" exploration. Try 5 things in sandbox before committing to a branch. Faster feedback, zero codebase pollution.
Comprehensive CSV data analysis and visualization tool. Use this skill when analyzing CSV files, generating data summaries, creating visualizations from data, detecting outliers, finding correlations, assessing data quality, or creating data reports. Triggers on CSV analysis, data exploration, data visualization, data profiling, statistical analysis, or data quality assessment requests.
Expert blueprint for Metroidvanias including ability-gated exploration (locks/keys), interconnected world design (backtracking with shortcuts), persistent state tracking (collectibles, boss defeats), room transitions (seamless loading), map systems (grid-based revelation), and ability versatility (combat + traversal). Use for exploration platformers or action-adventure games. Trigger keywords: metroidvania, ability_gating, interconnected_world, backtracking, map_system, persistent_state, room_transition, soft_locks.
Resolve implementation ambiguities before planning begins. Two modes: Discussion mode surfaces gray areas with concrete options for greenfield work. Assumptions mode reads the codebase, forms evidence-based opinions, and asks the user to correct only what's wrong (brownfield work). Use for "discuss ambiguities", "resolve gray areas", "clarify before planning", "assumptions mode", "what are the gray areas", "before we plan". Do NOT use for broad design exploration (use feature-design) or for planning itself (use feature-plan).
Dispatch independent subagents in parallel for unrelated problems spanning different subsystems. Use when 2+ failures have independent root causes, multiple subsystems are broken independently, or user requests concurrent investigation. Use for "parallel", "multiple failures", "independent bugs", "fix these concurrently". Do NOT use for related failures, shared-state problems, or exploratory debugging where root cause is unknown.
Facilitate structured idea exploration and product/design specification. Use when a user wants to talk through an idea, refine it via iterative questions, and converge on a clear design/spec (and later an implementation plan), especially after inspecting the current project state.
Trier un bug ou une issue en explorant le codebase pour trouver la cause racine, puis créer une issue GitLab ou GitHub avec un plan de correction basé sur le TDD. À utiliser quand l'utilisateur signale un bug, veut créer une issue, mentionne « triage » ou veut investiguer et planifier la correction d'un problème.
Use this skill whenever deciding what features to extract from raw marketplace assets — listing photos, owner-entered listing metadata, sitter wizard responses — to power item-to-item (similar listings), user-to-item (homefeed ranking), or user-to-user (mutual-fit matching) recommenders in a two-sided trust marketplace. Covers asset auditing, first-principles feature decomposition from the decision the user is making, vision-feature extraction (CLIP, room-type classification, amenity detection, aesthetic and quality scoring), listing text and metadata encoding (categoricals, multi-hot amenities, H3 geo-hashing, sentence-transformer description embeddings, structured pet triples), sitter wizard design (information-gain ordering, multiple-choice over free text, genuine skippability, hard constraint versus soft preference), derived-composition patterns for i2i / u2i / u2u (precomputed ANN shelves, multi-modal fusion, two-tower affinity, symmetric mutual-fit scoring, interpretable subscores), feature quality governance (single registry, training-serving parity, coverage and drift alarms, PII scrubbing, schema versioning), and incremental value proof (one feature at a time, ablation A/B, kill reviews, exploration slice, permanent feature-free baseline). Trigger even when the user does not explicitly say "feature engineering" but is asking how to get more signal out of listing photos, listing metadata, or the sitter onboarding wizard, or how to improve i2i / u2i / u2u quality without blindly ingesting a new model.
PR-backed and current-main optimization manual for `moonshotai/Kimi-K2*` and `moonshotai/Kimi-K2.5*` in SGLang. Use when Codex needs to recover, extend, or audit Kimi optimizations, including K2 router/MoE fast paths, K2 thinking Marlin paths, K2.5 wrapper/multimodal/runtime plumbing, W4AFP8/W4A16 quant tracks, parser contracts, LoRA coverage, and backend-specific validation.
Build high-quality visual Web artifacts using HTML/CSS/JavaScript/React — web pages, landing pages, dashboards, interactive prototypes, HTML slide decks, animated demos, UI mockups, data visualizations, and more. Use this skill whenever the user's request involves a visual, interactive, or front-end deliverable, including: - Creating web pages, landing pages, dashboards, marketing pages - Building interactive prototypes or UI mockups (with device frames) - Building HTML slide decks / presentations - Creating CSS/JS animations or timeline-driven animated demos - Turning design mockups, screenshots, or PRDs into interactive implementations - Data visualization (Chart.js / D3, etc.) - Design system / UI Kit exploration Even if the user doesn't explicitly say "HTML" or "web page," this skill applies whenever the intent is to produce something visual, interactive, or presentational. Not applicable: pure back-end logic, CLI tools, data-processing scripts, non-visual code tasks, command-line debugging.
This skill guides the use of Jupyter notebooks for data analysis, exploration, and visualization, particularly with BigQuery. It outlines best practices for notebook execution and validation (supporting both cell-by-cell execution and full notebook generation depending on tool availability), library installation, and structuring notebooks for clarity. It also covers specific rules for data cleaning, plotting, and integrating with BigQuery SQL and machine learning workflows. Relevant when any of the following conditions are true: 1. The user request involves a data analysis, data exploration, data visualization, or data insights task that requires multiple steps, queries, or visualizations to answer. 2. The user explicitly requests a notebook (.ipynb). 3. You are creating, editing, or executing cells in a Jupyter notebook. 4. You need to query BigQuery from within a notebook. DO NOT use the Python BigQuery client library; instead, you MUST use the `%%bqsql` magics explained in this skill.
Use when adding or updating Go CLI E2E coverage for one `tests/cli_e2e/{domain}` domain of the compiled `lark-cli`, especially when the work requires live `--help` or `schema` exploration, scenario-based `clie2e.RunCmd` workflows, and per-domain `coverage.md` maintenance.