Loading...
Loading...
Found 3,074 Skills
Deeply interviews the user about a bug before investigating or fixing it. Use this when the user says "interview me about [bug]", "I have a bug to report", "let's work on this bug", "found a bug", "there's a bug", or describes a bug they encountered. Asks probing questions about reproduction steps, environment, patterns, impact, and context that help isolate the root cause. Continues until the bug is well-understood, then writes a detailed investigation/fix plan.
Learn unfamiliar implementation patterns and fill in missing pieces when building features in a specific environment (e.g., data fetching in a particular runtime). Use when you need to discover or confirm patterns for an unimplemented area; implementation uses erudita and osgrep.
Create and manage Architecture Decision Records (ADRs). Use when documenting technology choices, design decisions, or architectural changes that need to be tracked over time. This is the CANONICAL ADR skill - all ADR-related work should use this skill.
Set up a project's meta-structure for agentic engineering — CLAUDE.md, AGENTS.md, docs/, README, and CHANGELOG. Use when starting a new project or retrofitting an existing codebase for the forge workflow.
Byzantine consensus voting for multi-agent decision making. Implements voting protocols, conflict resolution, and agreement algorithms for reaching consensus among multiple agents.
Document debugging sessions with hypothesis tracking and knowledge base
Create or update repository README.md files with concise, human-skimmable structure and verified metadata. Use when writing or refreshing README content, adding or correcting badges, documenting stack and deployment/services, and explicitly reporting testing coverage (unit, integration, e2e api, e2e web) plus CI execution status.
Create a new track with specification and phased implementation plan
Improves CLAUDE.md by analyzing conversation patterns to capture recurring preferences and corrections. Use when Claude keeps repeating a mistake, when you want to teach it a new preference, or when consolidating guidance from repeated instructions.
Use this when you need to execute R2 in the sdlc-dev product requirement Spec process, transcribe requirements/solution.md into a deliverable, acceptable, and testable requirements/prd.md, while avoiding guessing file paths, continuing generation when solution.md is missing, or using "Pending Questions/Open Questions" to replace the verification checklist.
Use when you need to execute I1 (Implementation Plan) in the Spec Pack of sdlc-dev, convert requirements/design into `{FEATURE_DIR}/implementation/plan.md` (the single source of truth (SSOT) for execution checklist and status), and provide an unambiguous task list for subsequent I2 execution.
Minimal TTS smoke test for Model Studio Qwen TTS.