impl

Implements one spec-driven project phase end-to-end with quality gates and independent code review.

37|17|Updated May 4, 2025
One-click install
npx skills add https://github.com/milvus-io/milvus-workshop --skill impl-milvus-io
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: impl
Source: https://github.com/milvus-io/milvus-workshop/tree/main/agent/.agents/skills/impl
Command: npx skills add https://github.com/milvus-io/milvus-workshop --skill impl-milvus-io

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a project implementation plan into production-quality code often fails because specs are vague, phases ship half-finished, and reviews get skipped. This Skill enforces a disciplined workflow: polish the relevant feature specs until they are implementation-ready, build exactly one phase of the impl plan, run strict quality gates, and complete an independent code review before declaring done. ## Core Features & Use Cases - Spec-first gating: Expands and polishes feature specs, diagrams, and the impl plan under ./specs until another engineer could implement the phase without guessing, before any production code is written. - Phase-scoped implementation: Works through the phase's task table in dependency order with typed, tested, benchmarked code and no TODOs, stubs, or gate bypasses. - Independent review loop: Spawns a review subagent against the polished specs, fixes valid in-phase findings, and records out-of-phase items in a deferred-findings backlog spec. - Use Case: A user says "build phase 2 entirely" — the Skill reads the impl plan and cited specs, polishes them, implements each task, runs ruff, mypy --strict, pytest, and security checks, then reviews the diff and reports exit criteria with evidence. ## Quick Start Use the impl skill to polish the relevant specs and build phase 2 of the impl plan end-to-end, then run the independent review.

Frequently Asked Questions about impl

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I implement a phase of a project impl plan with an AI agent?

Name the phase or milestone (for example "build phase 2" or "land M1") and the Skill reads the impl plan, polishes the cited specs, implements each task in dependency order, runs quality gates, and finishes with an independent code review before reporting done.

What happens if the specs are not detailed enough to implement?

The Skill treats spec polish as a hard gate and will not write production code until the relevant feature specs define behavior, APIs, invariants, error models, tests, and exit criteria. If a product or architecture decision is missing, it asks the user before coding.

Which quality gates does the impl workflow run?

It runs the project's documented gates: ruff format and lint checks, mypy --strict or equivalent type checking, the full pytest suite, plus bandit and dependency audits for boundary modules. Gates are never bypassed with suppressions or skipped tests.

When should I not use the impl skill?

Do not use it for one-off bug fixes, ad-hoc feature requests outside a planned phase, greenfield design, or throwaway prototypes. Use direct edits for small fixes and a spec-writing skill first when no impl plan or specs exist.

What happens to code review findings outside the current phase?

Valid out-of-phase findings are appended to a single deferred-findings backlog spec with severity, file:line citation, and a fix shape, so later phases can pick them up without inflating the current phase's scope.