kdlbs/kandev

ACP Debug runs an ACP agent CLI, speaks raw JSON-RPC, and records every wire frame to JSONL. Use it to probe capabilities, inspect handshakes, compare agents, or diagnose startup and session failures.

Add Integration scaffolds a new third-party service integration with per-workspace credentials, auth health polling, settings UI, and link/import controls. Use it when adding an external service like Jira or Linear.

Clean Branches removes local Git branches whose remote tracking refs are gone, including any associated worktrees. Use it after merges or PR cleanup to prune stale branches safely.

Code Review reviews changed code for quality, security, and architecture compliance when you explicitly ask for a local review or a PR finding needs investigation. It expects file-and-line findings with reasons and fixes.

Commit stages and creates git commits using Conventional Commits. Use it when there are dirty or staged changes, when asked to commit, or before pushing a PR.

Context Engineering curates the right project files, specs, and evidence before coding or debugging. Use it when starting a session, switching code areas, or reconciling external instructions with repository conventions.

Debug diagnoses Kandev bugs, running-instance issues, UI/browser failures, and runtime behavior. Use it to triage, gather evidence, and identify a root cause before handing off to /fix for code changes.

Docs Maintainer helps decide when Kandev's public docs need updates and applies them in the right place. Use it when CLI commands, config, workflows, APIs, or user-facing terms change.

E2E Tests writes and runs Playwright end-to-end tests with a red-green-refactor workflow. Use it to locate specs, run the right project, and debug failing browser tests.

Fix diagnoses a bug, updates the affected spec, and creates a reviewable fix plan with task files before any production code changes. It is used for durable bug repairs that will be implemented later with TDD.

Harness Improvement turns session learnings into durable harness changes like skills, agents, commands, and instruction files. Use it when updating Claude Code or related harness guidance across platforms.

Implementation Plan creates a committed plan and task files from a feature spec. Use it after writing a spec and before implementation to map codebase changes, tests, and execution order.

Interview Me clarifies the user’s intent before specs, plans, or code when an ask is underspecified. Use it to ask focused questions and surface assumptions before proceeding.

Kandev Plugin guides creating, fixing, testing, packaging, and publishing a Kandev runtime plugin in its own repository. Use it when the work targets a plugin implementation or release lifecycle, not the Kandev host or SDK.

Mobile Parity ensures Kandev UI changes preserve desktop and mobile capability while using native mobile interaction patterns. Use it for any user-facing UI change and to add mobile Playwright coverage.

PR creates a pull request from a committed, pushed branch after task checks pass. It uses the current host flow and can continue with fixup in the same conversation.

PR Fixup waits for PR checks and reviews, then fixes valid CI failures and primary-thread comments, verifies the result, and pushes updates. Use it when a pull request needs remediation before merge.

PR Walkthrough generates a single-file HTML summary of a pull request. Use it when you need a scrollable page that explains the problem, approach, code changes, data flow, and trade-offs for reviewers.

Planner Orchestration enforces Kandev's single-session workflow for planning, execution, delegation, and delivery. Use it to keep specs, task files, model checkpoints, and handoffs organized during feature, fix, debug, review, and verification work.

Playwright CLI automates browser interactions, page navigation, and screenshots from the command line. Use it to test web pages, inspect snapshots, and run Playwright-based workflows.

Product Demo Seeding builds truthful, disposable Kandev states for screenshots, product videos, and reproducible UI captures. Use it when you need believable workflows, agents, integrations, sessions, diffs, or native mobile states before capture.

Product Video Capture records, stages, and validates polished product films, landing-page loops, screenshots, and alternate framing from isolated demo data. Use it for product videos, GIF-like demos, cursor-follow camera motion, desktop/mobile captures, recaptures, and landing media.

Push verifies a committed branch is clean and that required task checks passed before sending it to its remote. Use it when the work is ready to publish, and with --fixup it continues into CI and review handling.

QA verifies a feature after implementation by tracing the wiring, running the happy path, and trying to break it with edge cases and failure modes. Use it when you need an actionable finding or explicit adversarial validation.

Record Knowledge keeps architectural decisions and product specs in sync with ongoing work. Use it when a request changes a durable boundary, contract, ownership rule, or repo-wide convention.