delivery plugins

by Agent Harness Skills contributors

Reusable skills for building agent-ready repository harnesses.

by Lev Nikolaevich

Independent plan and delivery reviews with evidence-based verdicts.

by Matt Corbett

Film & Video Production specialist team — agents (production-lead, line-producer, post-production-supervisor, production-finance-analyst), skills, templates, commands, an advisory hook, best-practice rules, a research-grounded knowledge bank with Mermaid decision trees, a scenarios bank, and a stdlib production-economics calculator (shoot-day cost, contingency burn, overtime burden). A production-management team for a producer, production company, or post house — it budgets to a defensible top-sheet, schedules to the shoot day rather than the calendar, runs the post pipeline as a dependency chain, and reads production economics the way a line producer who answers for every dollar on the day must. Inherits ravenclaude-core protocols.

by Wildcat Labs

Receipted delivery through merge and closure, with ranked frontier, audit, fuzzing, prose, specification, debugging, hardening, telemetry and measurement skills.

by Matt Corbett

Project & delivery management team — four specialists across the predictive (PMBOK/PMP) and agile (Scrum/Kanban) tracks and the hybrid between: delivery-lead (charter, schedule, scope/change control, earned value), scrum-master (backlog, sprints, ceremonies, velocity, impediments), risk-and-raid-analyst (scored qual+quant risk, RAID depth, mitigation/contingency, issue triage), and stakeholder-comms-lead (stakeholder register, comms plan, status/exec reporting, escalation memos, steering packs). Deepens — does NOT replace — ravenclaude-core's domain-neutral project-manager (the lightweight RAID/status-hygiene default every plugin routes to); this is the deep PM craft layer. Knowledge: a predictive-vs-agile-vs-hybrid decision tree + a best-practices library. Seams: prose polish → ravenclaude-core/documentarian; system design → architect. Requires ravenclaude-core@>=0.7.0.

by deep-code-review contributors

Perun — bring the thunder to your codebase. A portable, evidence-grounded quality bar for building with AI agents: the flagship deep-code-review skill (default install is review-only), plus opt-in overlays — gated delivery, an idea critic, a suite conductor, and stage-aware product, growth, positioning, business-ops, and output-safety specialists.

by Lukas Sliwka

Governed, spec-driven delivery for agent-built software: spec → operator authorization → governed build → an honestly-tiered merge floor → certification against the running app → human sign-off.

by 10Clouds

A shareable software-delivery workflow for Claude Code: ticket-driven build cycle with an explicit prototype-to-production rung, a one-question interview, verification on the live app, self code-review, an adversarial critic, and a session board so parallel sessions on one machine do not collide.

by Abdelazim Ali

Agentic Delivery Lifecycle: interview-driven repo intake, app classification, hardened Dockerfiles, Terraform on Azure, GitHub Actions CI/CD with immutable tags, deployment verification and Jira tracking. Every platform choice is an interview option.

by Khalil Maaouni

Turn an idea into a verified result with a guided start, honest forecasts, clear status, and a checked delivery. Verified on Claude Code; the plugin install is proven end to end by scripts/release-smoke-install.sh, and the tagged git clone remains the immutable option for auditors.

by Galileo Studio

Implement, review, and deliver an approved plan slice by slice through Worker and Reviewer subagents: a test gate on the Worker, a read-only Reviewer, and a human-approval prompt for commands that deploy, publish, change shared state, or destroy work.

by Job Schepers

The host-neutral delivery engine: 16 flow-* staged procedures, five of them one per devbook folder, 3 shared phases, and the pull-request lane, sequenced by the flow-runner agent through a closed set of extension points — spec, implement, verify, app.start, qa.run and deliver as services, session.start, flow.start, data.prepare, docs.update and flow.end as chores — with human gates a repository adds but never removes, and a run surface resolved from the live tool list so none is a normal outcome. Bindings, extensions, policy, and gates are declared per repository in .devbook/config.json.

by Rob Cherry

Delivery and collaboration discipline for fast builders and their coding agents, for Codex.

by Job Schepers

The unattended lane, stacked on delivery: the eleven schedule-* skills that run with nobody watching, plus install, plus the catalog of triggers that fires them. Nine schedulable entry points that pick their own input and run a flow or a review — bug fix, merge review, package update, performance review, review, security review, week starter, weekly cost analysis, and what's new — and six trigger definitions naming a target, a cadence, the plugins that target needs, and the prompt for a cloud session that starts with nothing but the repository. delivery-schedule:install creates or updates them through whatever scheduler the live session exposes and records the selection under components.schedule; schedule-status reads their runs and logs; schedule-run fires one now. One capability, two host names: Routines in Claude Code, Automations in the GitHub Copilot app. Depends on delivery and names devbook: a trigger whose target is not enabled is reported and skipped, never scheduled.

by Hieu Phung

An automation-first, Orca-supervised delivery protocol for coding agents.

by Job Schepers

Owns a repository's .devbook/config.json and explains the stack that reads it. `setup` writes the engine-owned keys — bindings, extensions, policy, gates — before anything is installed, then hands every component its own install skill; `update` moves the whole configured stack forward in one run, fanning out to every adopted component’s own install skill and skipping what this machine has not installed or this checkout has not enabled, migrations included; `guide` answers a question about this marketplace from what is on disk — what devbook, the delivery engine, the bridges, the surfaces, and the fan-out lane are and how they fit, which version of each plugin is installed against the newest published, and which are enabled; `adoption` reports where the .ai adoption record no longer matches what is installed and hands the write to flow-ai. All four are backed by a read-only report that names the file behind every fact. Declares no dependency, devbook included: it is named for the folder it writes into, not for a plugin it needs, and one it names but cannot find is reported as not installed.

by Job Schepers

The fan-out lane: a backlog turned into parallel work across sessions and worktrees — the one thing a flow may never do. fleet-issue-sweep triages a repository's open issues for relevance and for collision with work in flight, proposes the stale ones for closure, and dispatches up to five independent worker sessions, one issue and one worktree each, then waits them out and writes the brief. fleet-resolve-issue is what each worker runs: one issue to a pull request when the change proves itself, or to a parked worktree with a handoff brief when a human has to look. fleet-morning-brief re-reads a past sweep from its manifest and worker result files. An L1 extension over delivery: it owns no flow and holds no gate, and trades Personal Validation for a pull request nobody merges unread.

by Eduardo Furihata

Skills de entrega do Furihata — uma escada só: os alvos /work, /pull-request, /homolog e /prod levam o trabalho de onde estiver até o próprio estágio, cada um invocando o anterior quando falta algo atrás; os modificadores /repro e /card compõem com qualquer alvo, em qualquer ordem; /infra mapeia a infra do projeto; o processo mora em .claude/ship/setup.md, uma seção por skill. Funciona sozinho, sem depender de outro pacote.

by jayte

SDLC workflow plugin, one source for Claude Code and Codex: one `$wf` workflow skill (22 keys — the ten lifecycle stages intake through retro, the intake modes fix/rca/investigate/discover/audit/hotfix/refactor/update-deps/ideate/brainstorm/adopt, design, probe, simplify, auto, task, status/recap, close, ship-plan, docs, observability; `yolo` is Claude Code only) plus consult, imagery, uiproto, diataxis, and study-sources — over a shared runtime, hub, and .ai/ artifacts. Requires a hooks-GA Codex CLI (verified on 0.146.0) with hooks enabled.

by paulkim-lansik

Delivery-loop skills — land and release already-verified work through a repo's real gates. Tracker- and branch-model-agnostic: reads a consuming repo's own config instead of hardcoding one repo's setup.

by Andrey Myasnikov

Compose backlog scope and deliver selected Task Manager work with autonomous test fixtures, strategic explanations, and explicit critical-codebase acceptance.

by Nochevnyi Serhii

Artifact-driven delivery pipeline on top of GSD: deep investigation, ticket decomposition with a dependency graph, and per-ticket worktree->PR delivery babysat to green (CI + CodeRabbit/Copilot re-review). See docs/gsd_multilevel_delivery_pipeline.md.

by Abdelazim Ali

slipway implements the Agentic Delivery Lifecycle (ADLC): interview-driven repo intake, app classification, hardened Dockerfiles, Terraform on Azure, GitHub Actions CI/CD with immutable tags, deployment verification and Jira tracking. Every platform choice is an interview option.

by Software Factory contributors

Jstack delivery, architecture, evidence, review, and writing skills for production software work.