Bring out what your tools can do — bridge skills, MCP, and APIs into reusable agent, terminal, and automation workflows
Sumac exposes 5 tools with basic schemas and descriptions, but significant gaps limit production readiness. Three tools (get_available_skills, get_skill_details, get_skill_related_file) have minimal descriptions (20-80 chars) and lack parameter descriptions entirely. The remember_state and read_state tools are defined in a Python test fixture, not the main Rust server code, suggesting they may not be part of the primary implementation. Input schemas are present but sparse: most parameters lack descriptions, and no output schemas are documented. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are present despite clear risk classifications. Error handling is absent, no recovery guidance or actionable error messages visible in the code.
List available skills with their prompt, tool, and resource metadata
Get detailed information for a skill, including its prompt body and file listing
Read a file from within a skill directory using a safe relative path
Read a value from session memory.
Store a value in session memory.
Parameter descriptions missing across all tools. LLMs cannot infer what 'return_type', 'skill_name', 'relative_path', 'key', 'value' parameters control without explicit descriptions.
No output schemas documented. Callers cannot plan downstream tool calls or extract required fields (e.g., does get_available_skills return skill IDs for chaining to get_skill_details?).
Tool descriptions are too brief (20 - 80 chars). 'List available skills with their prompt, tool, and resource metadata' lacks WHEN to use this tool and what distinguishes it from get_skill_details.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications. LLMs cannot determine which tools are safe to retry or have side effects.
remember_state and read_state defined in test fixture (tests/fixtures/stateful_mcp_server.py), not main Rust server. Unclear if these are production tools or test-only stubs.