orassayag
UserAn ai agentic Claude Code workflow that takes a rough feature idea from a plan through review, finalization, boilerplate scaffolding, and parallel-agent execution - with built-in cross-project memory for lessons and preferences. Built in July 2026, it driven by 7 skills: /init, /plan-review, /plan-finalize, /boiler, /boiler-update, /orca, /lesson
Categories
Indexed Skills (37)
impeccable
Use when the user wants to design, redesign, shape, critique, audit, polish, clarify, distill, harden, optimize, adapt, animate, colorize, extract, or otherwise improve a frontend interface. Covers websites, landing pages, dashboards, product UI, app shells, components, forms, settings, onboarding, and empty states. Handles UX review, visual hierarchy, information architecture, cognitive load, accessibility, performance, responsive behavior, theming, anti-patterns, typography, fonts, spacing, layout, alignment, color, motion, micro-interactions, UX copy, error states, edge cases, i18n, and reusable design systems or tokens. Also use for bland designs that need to become bolder or more delightful, loud designs that should become quieter, live browser iteration on UI elements, or ambitious visual effects that should feel technically extraordinary. Not for backend-only or non-UI tasks.
improve-codebase-architecture
Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
prototype
Build a throwaway prototype to flesh out a design — a runnable terminal app for state/business-logic questions, or several radically different UI variations toggleable from one route.
to-issues
Break a plan, spec, or PRD into independently-grabbable issues on the project issue tracker using tracer-bullet vertical slices.
to-prd
Turn the current conversation into a PRD and publish it to the project issue tracker — no interview, just synthesis of what you've already discussed.
triage
Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.
visual-plan
Turn ordinary text plans into rich interactive visual plans with diagrams, file maps, annotated code, open questions, and UI/prototype review when useful.
visual-recap
Turn a PR, branch, commit, or git diff into an interactive visual recap with diagrams, file maps, API/schema summaries, annotated diffs, and focused review notes.
boiler-update
Triggered by the slash command /boiler-update <boilerplate-name|all> <description>. Updates a boilerplate's bank rule or skill, its cookbook, the bank index, and the on-disk skeleton in one pass, based on a description of what to change, add, remove, or restructure — keeping all four in sync so a convention change (e.g. "tests should live in __tests__ folders, not a root tests/ folder") takes effect the next time /boiler scaffolds from this boilerplate. `all` as the first argument applies the change to every boilerplate in the bank. If the scope is ambiguous (which boilerplate, or how many), asks for clarification before making any changes. Does NOT modify already-scaffolded live projects — that is always a separate, explicit request. Use whenever the user runs /boiler-update, mentions "update the boilerplate", "change the cookbook", "change a convention/rule for X boilerplate" or "for all boilerplates", "remove X from the template", "add Y to the boilerplate structure", or wants to evolve a boilerplate conve
boiler
Triggered by /boiler <path-to-plan>. Reads the finalized plan, selects the best-matching boilerplate using the built-in routing table, sets up local rules/skills from the bank BEFORE any copying or building, copies or builds the boilerplate, enriches CLAUDE.md, and commits the scaffold. Handles single and split fullstack projects. Use whenever the user runs /boiler, mentions "scaffold from boilerplate", or "start a new project from a template".
flow-compare
Triggered by /flow-compare <github-url-or-local-path> [path-to-reviewer-blocks] [--force]. Compares THIS workflow repo against another agentic-workflow repo and produces a structured gap report (seven lenses, severity emoji, gaps table + prioritized recommendations, fixed verdict) in docs/flow-compare/. Freshness-checked: if the target hasn't changed since the last run, it reports and stops at near-zero tokens. Use when the user runs /flow-compare, or asks to "compare our workflow against <repo>", "gap-compare with X", "what does <repo> do that we don't".
lesson
Triggered by "/lesson add <tag> <description>", "/lesson note <description>", or "/lesson-prune". Manual escape hatch for capturing a lesson outside a /orca run, a zero-friction inbox for unresolved observations (note mode), and periodic maintenance to flag stale lessons for deletion. Most lessons are captured automatically during /orca — see that skill's Phase 8a — this is only for the "I want to note this right now" case, or for bank upkeep.
master
Triggered by the slash command "/master path/to/plan.md". Executes a finalized plan sequentially: one Claude Code agent per human-sized stage, a review gate with a full diff + drafted commit message before every commit, no worktrees, branches, or parallel waves. Use instead of /orca in time-boxed, observed, or interview-like settings, or whenever the user wants to review every diff before it lands.
mode
Triggered by /mode, /mode default, or /mode god-mode. Reads or switches the current project's model mode without re-running /init: writes .claude/model-mode.json, updates the model field in the project's .claude/settings.json (merge, never clobber), and — if /boiler has already added a "## Model Policy" section to CLAUDE.md — updates that section in place so the human-readable copy stays in sync.
orca
Triggered by the slash command "/orca path/to/plan.md". Takes a finalized project plan and executes it in parallel: splits it into independent task chunks, spawns coding agents in cmux panes (default 2 per wave, hard ceiling 4 — both constants live in SKILL-lib.sh; every wave size is confirmed by the user before spawning; each agent works in its own git worktree and branch), monitors progress via status.md, rebases each completed branch onto the default branch, fast-forwards it with --ff-only, auto-resolves simple conflicts, escalates complex ones to the user, and cleans up worktrees and branches once every wave is integrated. Use this skill whenever the user runs "/orca", or asks to "execute the plan," "run the plan in parallel," "spawn agents on this," "kick off multi-agent execution," "parallelize the build," or otherwise wants a finalized plan carried out using multiple coordinated coding agents.
plan-finalize
Triggered by /plan-finalize <path-to-draft>. Reads the resolved draft produced by /plan-review, writes the final plan to docs/plans/ as a .md file, and deletes the draft. Run in a fresh terminal session after /plan-review completes.
plan-review
Triggered by /plan-review <path-to-plan> [path-to-reviews] [--auto-accept] or /init [god-mode]. Handles git init, CLAUDE.md, folder setup, and model-mode selection (/init), then reads a human plan and optional agent reviews, builds a unified issues table, runs the resolution loop, and writes the resolved draft to docs/pre-plans/. When done, tells the developer to start a new terminal session and run /plan-finalize. With --auto-accept, skips the resolution loop, accepts every recommendation, and runs the plan-finalize logic in the same session.
prove
Triggered by /prove [path-to-plan-or-spec]. Proves a build actually works, in inspectable artifacts — screenshots, captured requests, command outputs, before/after values — one per spec item, plus a human-eye manual checklist. User-invoked and deliberately heavy: run ONLY when the user asks ("prove it", "show me it works", "verify this for real", "/prove"), never automatically and NEVER as an orca stage. Recommended before submitting a home assignment.
recover-session
Triggered by /recover-session [name-or-keyword]. Recovers a dead or closed Claude session — an orca agent pane, the orca orchestrator, or a solo session — by finding its resume UUID + cwd and producing the exact `claude --resume` command. Use when a cmux pane died mid-run, a terminal was closed accidentally, or the user says "recover session", "resurrect <agent>", "bring back <task>", or "find the transcript for X".
scout
Triggered by /scout <area-or-topic>. Briefs the developer on an unfamiliar area of an EXISTING codebase before they write a plan: existing patterns with file:line pointers, conflicting mental models, impact map, watch-outs, and open questions — cached to docs/pre-plans/scout-findings.md, which /plan-review reads automatically. Use when the user says "scout the X area", "scout this", "give me a rundown of Y", "brief me on Z before I plan", or wants to understand code before planning. Does NOT propose implementations, write code, or pick an approach — that's the plan's job.
watch-pr
Triggered by /watch-pr [PR-number]. After pushing, waits for the PR's automated review + CI to land, then triages findings — auto-fixes mechanical ones, batches judgment calls into one question, replies-with-reasoning on disagreements. Hard cap of 2 rounds; terminates explicitly. ON-DEMAND ONLY — never part of the plan→boiler→orca flow (which is PR-less); exists for production/job work. Use when the user says "watch the PR", "wait for the review", or "handle the review when it comes".
comment-hell
Triggered by /comment-hell [paths]. Enforces the commenting guidelines on over-commented code — strips narrating comments so the code self-documents. Use when the user invokes /comment-hell, or says "too many comments", "clean up the comments", "comment hell". Scope: the current uncommitted diff, or the file paths passed as arguments.
stash
Triggered by /stash [hint]. Saves the current Claude Code session (project root + session ID + inferred alias + one-line note) to ~/.claude/stash.jsonl so it can be resumed later from any terminal via the stash helper. Use when the user says "/stash", "stash this session", "save this session for later", or "I need to close this window but resume later".
subagent-review
Triggered by /subagent-review [PR-number | branch | commit-range]. Spawns a senior expert sub-agent to review a PR, branch, or local diff and return a structured tradeoffs review (Pros, Cons, industry-standard check, non-standard solutions, verdict). On-demand only. Use when the user says "subagent review", "expert review", "have an X expert review this", or asks for an outside review of a change.
codebase-design
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
diagnosing-bugs
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.
domain-modeling
Build and sharpen a project's domain model. Use when the user wants to pin down domain terminology or a ubiquitous language, record an architectural decision, or when another skill needs to maintain the domain model.
modern-best-practice-react-components
Build clean, modern React components that apply common best practices and avoid common pitfalls like unnecessary state management or useEffect usage
use-modern-browser-apis
Utilize built-in browser APIs (like Popover API, View Transitions etc) instead of building features manually via JavaScript
tdd
Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.
implement
Implement a piece of work based on a PRD or set of issues.
clean-typescript
Write clean, efficient TypeScript code that follows common best practices
modern-accessible-html-jsx
Write clean, modern, and highly accessible HTML & JSX code, using semantically correct elements and attributes
modern-best-practice-nextjs
Build modern Next.js apps with App Router and best practices
modern-tailwind
Build clean, scalable UIs with Tailwind CSS using modern utilities and variants
resolving-merge-conflicts
Use when you need to resolve an in-progress git merge/rebase conflict.
web-security
Enforce web security and avoid security vulnerabilities
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.