YasMax91
UserClaude Code plugin for Laravel backends: grounded spec-driven development, blast-radius discovery before planning, and Definition-of-Done gates (format, static analysis, tests, OpenAPI).
Categories
Indexed Skills (14)
openapi-audit
Audit and repair a project's whole OpenAPI/Swagger spec — inventory every API route, find endpoints that are undocumented or documented partially, and fill the gaps from real code. Use for pre-existing spec debt (a project whose Swagger is empty or stale), not for the endpoints of the task you are currently implementing.
deep-discovery
Deep, multi-agent blast-radius discovery — fan out impact-mapper across every seed in parallel, dedup into one map, verify dynamic edges. Use ONLY for L3/L4 changes touching many seeds, by deliberate escalation from `start-task` discovery or by the user; drives a Workflow (~15× tokens). Escalates start-task discovery.
deep-grounding
Deep, multi-agent grounding of an external API — fan out cited readers across the docs, build a capability matrix, then an adversarial panel verifies every row. Use ONLY for L3/L4 integrations, by deliberate escalation from `ground-integration` or by the user; it drives a Workflow (~15× tokens). Escalates ground-integration.
ground-integration
Ground an external/third-party API integration before coding — fetch official docs, build a capability matrix, mark unknowns, and require sandbox proof. Use whenever a task touches an external service, SDK, webhook, or a payment/booking/messaging provider.
deep-review
Deep, multi-agent risk review — find risks across every dimension, then adversarially verify each finding before reporting. Use ONLY for L3/L4 diffs, by deliberate escalation from `risk-review` or by the user; drives a Workflow (~15× tokens). Escalates risk-review.
grill
A relentless interview that stress-tests a plan, decision, or idea until nothing is left silently assumed.
client-doc
Write a client-facing document for a non-technical client — the outcome, what is in and out of scope, an estimate in the agent's real build time, and what is needed from them (with their own time on a separate line). English canonical file plus a Russian mirror under docs/client/. Not a substitute for an implementation spec.
final-check
Run the final AI-SDD self-review before handing off backend work — run format, static analysis, and targeted tests, verify the Definition of Done, and produce the handoff summary.
frontend-handoff
Produce the frontend-developer handoff docs in ai/frontend after a backend feature or change is fully implemented and the gates are green. Writes a Russian instruction doc — what to build, when, how, why, where, plus the API contract (no frontend code) — updates the affected reference docs, then offers to commit. Use as the closing step of any implementation that changes the frontend-facing surface (endpoints, response shape, validation, auth, states, visibility).
implement-approved
Implement an already-approved spec/plan for a Laravel backend through existing architecture boundaries. Use only after the user has explicitly approved the plan or spec in the current conversation.
init
Onboard a repository to the Groundwork workflow — generate thin domain-only AGENTS.md/CLAUDE.md by grounded discovery, install Laravel Boost, and drop .groundwork.json. Use once per project; re-run to refresh from current code.
spec
Create or update an implementation spec under docs/specs/ for a backend feature, bug fix, API/contract change, migration, or risky domain change. Use after discovery, before implementation.
start-task
Start a backend task through the Groundwork AI-SDD discovery/plan workflow. Use at the beginning of any non-trivial feature, bug fix, or API/schema/permission/workflow/integration change, before writing code.
estimate
Answer "how long will this take?" in the agent's own measured build time. Reads the estimate ledger — this author's real active minutes per task — and reports a number that rests on a median and its sample size, with human time on its own line. Use whenever a duration, deadline, or delivery date is asked for or is about to be stated, in chat, in a plan, in a spec, or in a client document.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.