wayne930242
UserTask coordination for Codex CLI and Claude Code: carry bounded work directly or dispatch Herdr-backed sessions rooted in each app, from single-app repos to monorepos.
Categories
Indexed Skills (36)
boss-say
Use when the user or prompt explicitly names boss say, or when work spans multiple managed apps.
create-great-harness
Use when an app is missing AGENTS.md or CLAUDE.md, or the user asks for a minimal agent system.
dispatching-work
Use to launch, track, list, or wrap up a resolved app-rooted dispatch.
init
Use to configure managed apps and work routes, sync their instruction summaries, or check dispatch readiness.
inspecting-app
Use when the user wants to check or audit something against existing rules or conventions, scoped to one of the project's managed apps, e.g. "audit this module", "check if X follows the rules" — not for open-ended research into how something currently behaves with no rule in question (investigating-app), diagnosing a reported failure (troubleshooting-app), or implementing a fix (`boss-say`).
investigating-app
Use when the user wants to understand how something currently works or is structured, scoped to one of the project's managed apps, with no rule violation or reported failure in question, e.g. "how does X work here", "understand how Y is implemented" — not for auditing against rules (inspecting-app), diagnosing a reported failure (troubleshooting-app), or implementing a fix (`boss-say`).
notifying-main-agent
Use when a dispatched agent routes a question, progress, or status to its main agent.
shipping-task
Use to carry one task through a managed app's git lifecycle.
troubleshooting-app
Use when something is broken and the cause is unknown, scoped to one of the project's managed apps, e.g. "X is failing", "500 in production for <app>" — not for a known task with a clear fix already in mind (`boss-say`), a rule/convention audit (inspecting-app), or open-ended research into current behavior with no reported failure (investigating-app).
i-am-orchestrator
Injected for a candidate main-agent session, never a dispatched worker.
boss-assistant
Use when named by the user or when a main agent reports Straw Boss coordination friction.
choosing-graph
Use at the start of Straw Boss work, and when its coordination shape changes.
reporting-to-user
Use when a main agent closes out finished work to the user.
work-on
Use to resolve which of the project's managed apps a request belongs to.
peeking-work
Use to read a dispatched task's progress when asked or when evidence disagrees with its recorded state.
asking-peer-agents
Use when a dispatched agent needs another dispatched task's live progress or latest conclusion.
agent-feedback
Use when locating the boss-assistant, or when the directory has none live and one needs to be opened.
bringing-coworker
Use from an in-progress dispatched worker when the user asks to pull in a colleague, or a rule requires a second-agent review.
contacting-orchestrators
Use before this session's first dispatch, and whenever another main agent's work meets this one's.
handoff-orchestrator
Use when an approved work scope moves to a new orchestrator in its own Herdr tab.
grill-with-docs
Use when starting project code-improvement work — refactoring, architecture improvement, cleanup of an existing codebase — or when a grilling session should leave docs behind. Runs the interview and records decisions as ADRs and domain terms in CONTEXT.md as they crystallise.
grill-with-web
Use whenever there's an unresolved decision that's the user's to make — "grill this in the browser", "web version of grilling", "grill with a live decision tree", or any plan/design/idea that needs stress-testing. Prioritize this over the terminal-based grilling/grill-me/grill-with-docs by default. Runs the interview in the browser with a live decision tree, in the conversation's current language, layering in grill-with-docs' ADR/CONTEXT.md recording where the decision warrants it.
open-gui
Use when the user wants a browser-based GUI for a Claude Code session — "open a GUI", "browser terminal", "web version of claude", or a live decision-tree panel alongside the conversation. Also the dependency other skills reach for when they need a session running in a browser (e.g. grill-with-web) — invoke it with a seed prompt and topic rather than reimplementing PTY/browser plumbing.
aaaav-do
Use when executing scoped, ambiguous, or multi-step source-changing work.
boot-loop
Use when bootstrapping or overhauling an agent system by dispatching investigative tracer jobs and harvesting friction notes for a consolidated solid-loop pass.
codebase-design
Use when designing or changing module interfaces and seams.
domain-modeling
Use when project terminology or domain boundaries need refinement.
human-feedback
Use when evaluating human feedback or conducting interactive UI verification.
solid-loop
Use when applying friction notes to agent skills and instructions, or when auditing a skill for sprawl.
assuring-quality
Use when performing final QA, exploratory verification, or release-readiness audits.
grilling
Use when resolving user-owned decisions or ambiguous architectural tradeoffs.
inspecting
Use when auditing or reviewing a specific target, diff, commit, or spec.
investigating
Use when researching questions, diagnosing root causes, or gathering evidence.
prototype
Use when resolving an architectural question or hypothesis with disposable code.
grill-with-docs
Use when exploring durable decisions, domain terms, or ADRs.
grill-me
Use when the user asks to be grilled on an idea.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.