Agent Tool Contracts
Design agent/CLI tool schemas, naming, errors, and idempotency for LLM callers.
Promosso
Cosa fa
Provides concrete rules and a review checklist for designing tools, parameter schemas, and CLI surfaces meant to be called by LLM agents: verb-noun naming, minimal required parameters, explicit side-effect wording, bounded/paginated output, stable {code,message} errors, and an idempotency pattern table for read/create/update/delete. Triggers when the user is defining tools for agents, SDKs, MCP servers, or AI-native CLIs.
Rapporto di test
Applying its idempotency table to a `delete_project` tool turned a bare 'delete and return success' baseline into a two-step confirm + soft-delete design with a stable NOT_FOUND/CONFIRMATION_REQUIRED error schema — a concrete, checklist-driven improvement over default output.
Testato il: 2026-07-15 · Claude Code 2.x (agent harness)
Installazione
git clone https://github.com/charlieviettq/awesome-agent-skill mkdir -p ~/.claude/skills cp -r awesome-agent-skill/.claude/skills/agent-tool-contracts ~/.claude/skills/agent-tool-contracts
Comandi e prompt di esempio
/agent-tool-contractsDesign agent/CLI tool schemas, naming, errors, and idempotency for LLM callers.
Gli skill si attivano con richieste in linguaggio naturale, senza comandi da ricordare. Dopo l'installazione, prompt come questi lo attivano (in inglese):
Help me design the parameter schema for a new MCP tool that creates PRsReview this CLI's argument names and exit codes for AI agent compatibilityWhat's the best idempotency pattern for an agent-facing delete tool?