LastPM v3 — Python orchestration layer for the LastPM agentic PM system. Reads 73 SKILL.md files, registers each agent as an MCP tool with typed JSON Schema, routes through Python engine, and streams artifact generation back to the client.
LastPM MCP server has critical definition gaps. Tool 1 (lastpm_vault_status) has a minimal description (73 chars) and incomplete schema, product_name is documented but no return schema is specified. Tool 2 (dynamic_agent_tools) is problematic: it claims to register 73 tools dynamically, but the source shows only a description template, not actual tool definitions. The code loads agents via load_all_agents() and builds schemas via _build_tool_schema(), but we cannot verify the actual agent definitions, parameter types, or descriptions from the provided source. Additionally, both tools lack output schema documentation, error handling guidance, and parameter descriptions are minimal. The dynamic tool registration pattern is opaque, we see schema building logic but not the actual agent SKILL.md files that define the 73 tools.
73 dynamically registered agent tools loaded from SKILL.md files. Each agent is registered as a tool with the agent's purpose, framework, domain, and scope in the description. Input schema is built from the agent's input_schema with typed JSON properties.
Show which canvas fields are populated vs UNDEFINED in the current vault.
Tool definitions are inferred, not directly visible. The code loads 73 agents from SKILL.md files via load_all_agents(), but the actual agent definitions, parameter schemas, and descriptions are not shown in the source.
No output schemas documented. Neither tool specifies what fields or structure the response contains. LLMs cannot plan downstream calls or extract data without knowing the response shape.
lastpm_vault_status description is 73 characters, below the 10 - 1024 baseline and lacks WHEN/WHY context. It does not explain when to call this vs other tools or what the output format is.
dynamic_agent_tools schema is incomplete. The 'required' array is empty even though product_name is documented as auto-detected if only one product exists, unclear if it is truly optional or if the schema is incomplete.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | <=2025-11-25 | v2 |
No error handling guidance. Neither tool describes what happens on failure (e.g., product not found, API key missing, agent not found). The code shows error returns but no recovery hints for the LLM.