Grade a system prompt against inline test cases. Uses 6 criteria types (output_match, schema_validation, tool_usage, safety_check, custom_script, llm_judge). Returns score 0-1. Uses YOUR API keys.
| Name | Required | Description | Default |
|---|---|---|---|
| model | No | Model ID. Default: provider default | |
| provider | No | LLM provider. Default: anthropic | |
| test_cases | Yes | Test cases: { name, input, expected } | |
| system_prompt | Yes | The system prompt to evaluate |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint and openWorldHint. The description adds key behavioral context beyond these hints: it uses the user's own API keys, applies six specific criteria types, and returns a 0-1 score. This alerts the agent to potential costs and side effects without contradicting the safe-read annotation.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is three sentences, front-loaded with the primary purpose. Each sentence earns its place: purpose, criteria types, return value and API key caveat. No repetition or filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no output schema, the description appropriately discloses the score range and criteria list. It does not detail a complete response breakdown, but for a read-only inline grading tool with strong annotations and schema coverage, this is adequate. It could mention more about error handling or rate limits, but that is not essential.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 100% description coverage for all four parameters, so the baseline is 3. The description mentions criteria types but does not add further parameter-level detail beyond what the schema already provides. It does not clarify relationships between test_cases structure and the listed criteria, leaving the schema to carry the full burden.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses a specific verb ('Grade') and resource ('system prompt against inline test cases') and enumerates six criteria types, clearly distinguishing this as an ad-hoc inline grader rather than the suite-based kopern_run_grading. The scope is clear and unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The phrase 'inline test cases' implies use for quick, ad-hoc evaluation, but the description does not explicitly name alternatives or provide when-not guidance (e.g., when to use kopern_run_grading or kopern_create_grading_suite). No exclusions are stated, so guidance remains implied rather than explicit.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Add one secure layer between your agents and this server.
Most tools have clearly distinct purposes (create vs list vs get vs run vs connect). A few potential overlaps exist (deploy_template vs create_agent, import_agent vs create_agent, grade_prompt vs run_grading) but descriptions clarify the differences.
All tools share the 'kopern_' prefix and mostly follow a verb_noun pattern (create_*, get_*, list_*, run_*, connect_*). The exception is 'kopern_compliance_report', which uses a noun phrase without a verb, breaking the otherwise consistent naming.
With 31 tools, this exceeds the 25-tool threshold for well-scoped servers. While the domain is broad (agent lifecycle, grading, pipelines, teams, connectors), the sheer number of tools feels heavy and could be consolidated (e.g., a single 'manage_memory' tool already bundles multiple actions).
Core agent management (create, read, update, delete, list) is solid, and grading has suite creation, execution, and results. However, pipelines and teams lack get/update/delete operations, connectors only support connect (no disconnect), and there's no way to manage grading suites beyond creation and running. This leaves notable gaps for secondary resources.
