What problem does it solve? Projecting PoP's six canonical specialist roles into Kimi Code agent Markdown and configuration is error-prone when done by hand: tool allowlists, model routing, and the canonical body can drift or lose guarantees. This Skill generates deterministic agent files and candidate configs, then validates them locally without ever invoking Kimi Code, a model, or the network. ## Core Features & Use Cases - Deterministic agent generation: Builds Kimi Code agent Markdown from .agents/agents/<role>.md, preserving the complete canonical body with a SHA-256 source marker and fail-closed tool/subagent policies per role. - Symbolic model routing: Assigns primary routing to Planner and Judge roles and secondary routing to recon, orchestrator, executor, and verifier roles, resolving secondary through a [secondary_model] candidate config fixed to kimi-for-coding (K2.7). - Fail-closed local validation: Parses candidate TOML with tomllib, rejects unknown frontmatter fields, undemonstrated guarantees (effective effort, max nesting, aggregate validation), and K2.7 medium-effort claims, exiting BLOCKED on any violation. - Use Case: After editing the canonical pop-executor role, run the build command to regenerate its Kimi Code agent with secondary routing and a candidate config, then run validate to confirm the projection matches the source byte-for-byte before committing. ## Quick Start Ask the agent to project the pop-executor role into a Kimi Code agent with secondary routing and validate the result against the canonical source.
