Replit Agent Customization Reference

Authoritative reference for customizing Replit Agent: replit.md instructions, Agent Skills, Plan/Build modes, MCP client/server, and cloud-IDE constraints.

Sby Skills Guide Bot
DevelopmentIntermediate
008/5/2026
#replit#agent#mcp#skills#cloud-ide

Recommended for


name: g-skl-platform-replit description: Authoritative reference for Replit Agent (cloud IDE) customization in gald3r projects. Covers replit.md Agent instructions, Agent Skills in /.agents/skills, Plan/Build modes + effort tiers, MCP (client + hosted server), and cloud-IDE constraints (no hooks, no commands). crawl_max_age_days: 14 vault_doc_path: research/platforms/replit/ vault_docs_url: https://docs.replit.com/replitai/replit-dot-md docs_url: https://docs.replit.com/replitai/replit-dot-md docs_url_secondary:

  • https://docs.replit.com/replitai/skills
  • https://docs.replit.com/replitai/agent
  • https://docs.replit.com/learn/model-context-protocol
  • https://docs.replit.com/references/agent/task-lifecycle
  • https://blog.replit.com/introducing-workflows last_doc_scan: 2026-06-02 capability_status: hooks: "❌ no native lifecycle hooks; task lifecycle is observational; no hook surface for g-hk-*.py to wire into" rules: "✅ replit.md single instruction/memory blob (auto-read, self-updated); no .mdc / no glob scoping" skills: "✅ Agent Skills (agentskills.io SKILL.md) in /.agents/skills — lazy-loaded, Project/User/Enterprise scopes" commands: "❌ no user-authored slash-command registry; Workflows are shell-command runners, not agent commands" agents: "⚠️ native Plan/Build modes + effort tiers; no user-definable custom-agent file format" mcp: "✅ native — Agent is an MCP client (cloud UI); Replit also ships a hosted MCP server" token_budget: low subsystem_memberships: [PLATFORM_INTEGRATION]

HELP CONTRACT (T442 — cross-platform, non-substitutable)

If the invoking command's arguments are EXACTLY -h, --help, or help (one token, nothing else): do NOT run any operation of this skill. Respond ONLY with a compact usage card — the command's name, its one-line purpose, each documented argument/option on its own line (or "none"), and the path to its command file — then STOP. Read-only: no .gald3r/ writes, no state changes, no task/bug creation. This block lives in the SKILL (not a rule) because skills are the execution layer on every supported platform; rules are optional context on most.

Provisioned by gald3r platform install. Documentation: https://docs.gald3r.ai

Related skills