最近新增
刚加入收藏的新技能，抢先试用。
shipRelease
jthorare
"Ship a release — sync branch, run tests, audit coverage, push, open PR, then auto-chain → /landAndDeploy → /canary → /syncDocs (skip the deploy chain with --no-deploy)."
rootCause
jthorare
4-phase systematic debugging — investigate, analyze, hypothesize, implement. Auto-freezes scope to the module boundary to prevent scope creep.
review
jthorare
Orchestrate a multi-agent PR code review. Spawns domain-specific reviewer subagents in parallel based on changed files. Findings are saved to ~/.agentic-workflow/<repo-slug>/reviews/<pr>.json — run /postReview to publish to GitHub.
reflect
jthorare
After a piece of work lands, review how it went with three reviewers (judgment, tooling, divergent) and a synthesizer, and turn what they find into typed, evidence-backed proposals. Never edits files. Port of pstack's reflect (MIT).
productReview
jthorare
"Founder/product lens review of plans with 4 scope modes -- mvp, growth, scale, pivot. Challenges assumptions and tightens scope."
prismStatus
jthorare
"Health check for prism-mcp. Reports dashboard reachability, MCP server connection, knowledge base population, and version."
postReview
jthorare
Publish a completed /review to GitHub as batched PR comments. Reads from ~/.agentic-workflow/<repo-slug>/reviews/<pr>.json and posts one review per agent (minimizing API calls). Use after /review has written a local state file and you are ready to publish.
planDevexReview
jthorare
"Developer-experience review of a plan document. Maps friction in API ergonomics, setup, error messages, observability, docs, migration paths, and ramp-up cost."
planDesignReview
jthorare
"Design-lens review of a plan document. Rates information architecture, interaction states, user flows, accessibility, and brand/voice consistency on a 1–5 scale with specific gaps."
officeHours
jthorare
"Spec-driven brainstorming session with EARS-format requirements. Outputs plan.md (the canonical handoff) plus domain-specific docs (product.md, engineering.md, design-brief.md, TASKS.md) to plans/ directory — each assignable to its owning team."
new-slice
jthorare
Add one use case (a slice) to an existing C# package, red first — the failing test that names the behavior, request and response records, a stub service, validator, registration, CLI command or endpoint, typed ids and ports with their fakes.
new-package
jthorare
Scaffold a new C# package (Core + App, or Core, Features, Infrastructure, Host) that passes every check on its first commit and leaves exactly one red test — the TDD start for a port or a new tier-1..3 package.
meta
jthorare
Run a meta orchestrator over a team of lane sessions through the aw-meta CLI — replies owed, the lane map, gate and merge, owner questions, health and the running log — so nothing is dropped or typed by hand.
landAndDeploy
jthorare
"Wait for PR merge, run deploy command, poll health, run smoke tests, then auto-chain /canary. Configured by .agentic-workflow/deploy.json."
judge
jthorare
Tune, inspect, and undo cheap-agent-harness judge decisions in-session — config get/set, why, undo, health.
enhancePrompt
jthorare
Use when the user invokes /enhancePrompt — discovers available project documentation, reads relevant files, and rewrites the user's request with richer context before execution
design-verify-web
jthorare
Playwright screenshots at mobile/tablet/desktop viewports, diff against mockup baseline using design-comparison MCP, and report discrepancies with fix suggestions.
design-verify-ios
jthorare
Boot simulator if needed, capture screenshot via XcodeBuildMCP, diff against mockup baseline using design-comparison MCP. Reports discrepancies with fix suggestions.
design-verify
jthorare
Detect web vs iOS automatically and delegate to /design-verify-web (Playwright screenshots) or /design-verify-ios (XcodeBuildMCP screenshots). Diffs against mockup baseline.
tripo
dcramer
Make, texture, rig, animate or export 3D models with Tripo on the team's Studio subscription, through tools/tripo/tripo.ts. Use when asked to generate a model, concept image or views, retexture, retopologise, segment, rig or animate a model, or check Tripo credits.
review-desk
dcramer
Raise a question, a still or a film for the developer's review on their review desk (a claude.ai artifact page), read their answers, and turn settled design answers into decisions. Use when you need the developer's call or eye, when an agent hands you a FOR REVIEW line, when reading answers, or when setting up a desk the first time.
models
dcramer
Make a model for Terra that is not a character: a prop, station, furniture, tool or weapon, an item, a tree, plant or rock, from a description to a checked model in the game. Covers deciding its size, budget and materials; the concept in the art style (textured, lightly stylised realism); the Tripo model on the Studio subscription; the Blender clean-up that keeps its textures; putting it in the game's data; and checking it against the style guide. Use when asked to make, generate, add or replace a model for the game, or when adding something to the game that needs a model.
item-icons
dcramer
Make, check or fix item icons (what the action bar, backpack, equipment slots, storage and station screens show for an item) in Terra's painted style. Use when adding an item or changing its look (a held tool or weapon, a worn thing, or a material), when the icons test fails, when asked about item icons, or when an icon looks wrong or is missing.
environment
dcramer
Build or judge how a place looks in Terra: a biome's forest, ground cover, trees, light, shade, haze and time of day, against a measured reference of the look aimed at (another game's frames, photographs). Covers making the reference, the views and films a look is judged on, measuring before tuning, the checks every change passes (popping, collision, frame time), the adversarial review at milestones, and the traps that cost rounds before. Use when asked to make a biome or environment look better or like a game or place, to tune light, shade, fog or night, or to review how a place looks.