Governed AI Delivery (GovKit) is an open source toolkit for spec-driven, evaluation-aware AI-assisted software delivery.
This repository also configures its own agents. See what governed-ai-delivery tells them →
Accelerated-Innovation/governed-ai-delivery
Skill Claude Code
Given an arbitrary project, inspect it and produce a .checks/checks.d/ configuration so the project can run ./check / ./check-llm. This skill does not assume any language, package manager, or toolchain, and it does not install a pre-built "profile" — every check is derived from evidence found in the target repo.
Accelerated-Innovation/governed-ai-delivery
Skill Claude Code
Plan implementation work as a sequence of the smallest independently demonstrable increments. Use when given a feature request, a body of work, or any implementation task that must be broken down or sequenced BEFORE coding — it classifies the request, identifies the next demonstrable behaviors, and recommends…
Accelerated-Innovation/governed-ai-delivery
Skill Claude Code
Write or improve unit tests following a strict determinism, F.I.R.S.T., and fast-feedback-budget discipline. Use when asked to write tests, improve tests, review test quality, classify tests, or assess a test suite against a 30-second fast-feedback budget. Invokes a structured output protocol that identifies behaviors…
Accelerated-Innovation/governed-ai-delivery
Skill Claude Code
Standards for writing, reviewing, and evaluating unit tests. Use whenever writing new tests, reviewing existing tests, deciding what to test, assigning a test to a fast vs. full feedback loop, or diagnosing flaky, slow, or low-value tests. Covers structural fidelity, the FIRST properties, behavioral invariants, and…
Accelerated-Innovation/governed-ai-delivery
Skill Claude Code
Zero-One-Many Representation Evolution — use this skill whenever working with existing code that needs to change, grow, or be refactored. Triggers include: "refactor this", "add a new X to this", "this is getting messy", "how should I represent this", "there's a lot of duplication here", or any time numbered…
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Author an Architecture Decision Record for a new pattern, exception, or boundary change. Use when the user asks to write an ADR or invokes /govkit-adr-author.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Validate architecture boundaries, standards alignment, and ADR need before planning a feature. Use when starting a new feature or invoking /govkit-architecture-preflight.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Plan a provider-neutral evaluation suite for an LLM feature. Use when the user asks to plan model evaluations or invokes /govkit-eval-suite-planning.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Author a fix record for a defect that restores already-established behavior. Use when the user reports a bug, asks to fix a defect, or invokes /govkit-fix-record.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Validate Level 5 provider-neutral LLM architecture decisions before planning. Use after /govkit-architecture-preflight for features with mode:llm or when invoking /govkit-genai-preflight.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Generate an ordered implementation checklist with evaluation compliance summary from a validated preflight. Use when the user asks to draft implementation steps or invokes /govkit-implementation-plan.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Design a governed agent topology and produce agenttopology.md. Use for multi-agent features, delegation, adaptive orchestration, or when invoking /govkit-multi-agent-design.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Generate a feature plan (plan.md) and evalcriteria.yaml from NFRs and acceptance scenarios. Use when the user asks to plan a feature or invokes /govkit-spec-planning.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Author and open a governed pull request that carries its governance evidence. Use when the user asks to open a PR, create a pull request, or invokes /govkit-pr-author.
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Create honest atomic Git commits by choosing one coherent work batch before editing, keeping the whole repository green, resolving every untracked file, staging the entire root with git add ., and pausing for human review before committing. Use when starting commit-sized work, completing a demonstrable implementation…
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Diagnose and improve names for code objects — variables, functions, methods, classes, modules, and parameters — using Ottinger's naming guide. Use when asked to review, choose, or rename identifiers; when names are vague, lying, redundant, or noisy; when planning an incremental rename refactor; or when extraction…
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Establish trustworthy feedback around poorly understood or weakly tested existing code before changing it. Use for characterization or approval tests, test-harness access, seams, sensing and separation, dependency breaking, and incremental replacement of risky live code. Once a fast trustworthy boundary exists, use…
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Review code and propose evidence-based representation improvements using Tim Ottinger's Eight Code Virtues, SPOT, ZOM, and the improvement test. Use for broad craft, refactoring-readiness, clean-code, knowledge-representation critique, or discovering data clusters, data classes, and class-splitting boundaries. Do not…
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Split oversized features, epics, and stories into thin, demonstrable, deployable slices through progressive admission: start with a closed end-to-end skeleton, then admit one case at a time. Use to choose or sequence delivery slices, including message, schema, import, API, form, path, rule, persona, or channel…
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Implement or extend production behavior test-first, or diagnose focused unit-test quality and flakiness, using red-green-refactor, ZOMBIES, Tidy First?, FIRST microtests, and Save Your Game checkpoints. Use for TDD, microtests, test-first bug fixes, flaky tests, and refactoring within an active green cycle. Do not use…
Accelerated-Innovation/governed-ai-delivery
Skill Claude CodeCodex
Format thin vertical slices as explicit "User invokes" and "User uses result" outcomes. Use when the split already exists or the user specifically wants user-observable wording for slices. For choosing the split sequence itself, prefer story-splitting-for-delivery (progressive admission). Also use when reframing a…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: