Skill UI

This skill enables running the Pi agent non-interactively for repository analysis and implementation tasks. It involves setting the Bash working directory, checking Pi availability, and executing sessionless tasks using JSON mode. Users should delegate analysis or implementation to Pi by passing prompts as quoted arguments. Security measures include limiting tool access and avoiding credential handling.

This skill applies a rigorous, evidence-first discipline to coding tasks, inspired by Fable-like best practices. It guides agents through a structured workflow: inspecting the workspace, tracking goals and findings, grounding all conclusions in verifiable evidence, and executing thorough verification gates before declaring completion. Ideal for complex development, debugging, or adapting cross-model prompt guidance into robust, verifiable project rules.

This skill enables launching OpenAI's Codex CLI as an isolated subagent for bounded coding, review, or verification tasks. It runs non-interactively in a sandbox, streaming progress while keeping context separate. Ideal for parallel work, independent verification, and self-contained coding tasks with clear success criteria. Users must ensure authentication and handle file ownership explicitly.

Cohesivity provides headless backend services for AI agents via a single HTTP API, including hosting, databases, storage, and LLM access. It supports ephemeral tenants with 72-hour expiration, requiring user approval for new backends. Ideal for bootstrapping new apps or adding services mid-build without vendor consoles.

Turns durable, non-obvious findings into atomic, interlinked markdown notes with a maintained index and wiki-links, so agents retain knowledge across sessions instead of relearning facts.

This skill guides developers in building flexible and maintainable React components by applying composition patterns. It helps avoid boolean prop proliferation through compound components, state lifting, and internal composition. Suitable for refactoring, library building, and API design to improve code scalability for humans and AI agents.

Orchestrate multi-dimensional code reviews covering quality, security, performance, and architecture using specialized AI agents. Integrates static analysis, vulnerability scanning, and testing verification to provide actionable remediation guidance for complex software projects.

Build AI agents that interact with computers like humans, performing tasks such as viewing screens, moving cursors, clicking buttons, and typing text. This skill covers Anthropic's Computer Use, OpenAI's Operator, and open-source alternatives. It emphasizes sandboxing, security, and handling vision-based control challenges for desktop automation and GUI interaction.

Configure a Rails project to run inside Conductor workspaces for parallel coding agents: creates conductor.json, bin/conductor-setup, and script/server scaffolding, links shared dev secrets, and updates Redis config to use REDIS_URL with isolated ports and namespaces.

This skill acts as a context agent ensuring continuity between AI coding sessions. It automatically saves session summaries, decisions, and pending tasks to local storage and MEMORY.md. Upon starting a new session, it loads a detailed briefing to restore context. Features include full-text search, archiving, and maintenance tools for managing long-term project memory within the Claude Code environment.

Strategies for compressing long agent conversation histories using structured summaries, artifact tracking, compression triggers, and probe-based evaluation to optimize tokens per task.

This skill covers predictable performance degradation patterns in language models as context length increases, including lost-in-middle effects, context poisoning, and distraction. It helps diagnose agent failures and design resilient systems for large contexts.