These skills help an agent plan work before implementation and keep the loop visible while it changes. Members cover research, design, implementation, reviews, roadmap planning, launch work, and product office hours.
61 skills
25 engineering workflows for requirements, planning, implementation, testing, review, performance and production delivery.
HumanLayer workflows for visual explanations and staged research, design and implementation.
Run an iterative worker-reviewer cycle for code, specifications or writing, with actionable feedback and a bounded review loop.
Organize project rules, task context and references so coding agents receive relevant information at the right time.
Review AI engineering, learning and product questions through software paradigms, building to understand and reliability engineering.
Draft human-readable usage terms and a machine-readable manifest for personal materials.
Keep final artifacts, labels, commits, and handoffs focused on accepted requirements after a conversation includes corrections or discarded proposals.
Audit project rules, identify their source of truth, standardize naming, and build bridges for consistent use across agent hosts.
Create a source-based personal profile from messages and documents, with separate workflows for usage terms and handover protection.
Turn authorized messages, documents and other records into a person-profile Skill organized by work, interaction, memories and values.
Design and implement MCP servers with discoverable tools, clear schemas and task-based evaluations.
Identify repeated instructions in local coding sessions using a model-free report before deciding whether to create a profile.
Turn a video brief into a Claude Design prompt and composition, guided by an active personal video profile.
Create a person-perspective Skill from supplied documents, chat logs, audio or video; includes preset persona prompts.
Map incoming engineering work to the relevant workflows in Addy's Agent Skills collection.
Design agents around capabilities, domain knowledge, context and progressively added complexity.
Mine a personal working profile from AI coding sessions and use focused workflows for profile setup, repeated instructions and video work.
Review a specification or implementation plan with a multi-agent panel before starting work.
Capture command failures, corrections, missing capabilities, and better approaches as structured learnings that can improve future agent work.
Examine AI research directions, generalization, scaling and the relationship between capabilities and safety.
Maintain a durable workspace for complex reasoning and repository tasks, with explicit goals, evidence, decisions, and resumable progress.
Use language-specific Claude API and SDK references for streaming, tools and agent integrations.
Consolidate previous agent sessions into reviewed memory and skill improvements through an offline sleep workflow.
Prepare canary markers and optional protective measures for an authorized handover bundle.
Create a personal profile from your own session messages with dated evidence, preflight estimates and explicit approval before mining.
Pick a planning skill for the first pass, a review-loop skill after implementation, and a launch or roadmap skill when the work spans several milestones.