Menu
Use when adding or changing a skill, fragment, hook, or runtime file in this repo — including deciding which unit a new capability should be.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when you want to learn how experts would think about a design tradeoff, architecture choice, repeated failure, or domain question — including when you are stuck rather than asking, or the question is one of style.
Locks the session into DELEGATOR mode — route all work to subagents and workflows, do no work directly.
Use when a pipeline stage or the human meets facts that live outside the repo — a service, database, queue, third party, or live logs — and the work depends on their current state. Repo-local work skips it.
Turn rough ideas into clear work orders before planning or building. Use on ambiguous or underspecified requests, including ones that sound specific but name no outcome.
Use as load-bearing work lands, or when the user asks to debrief, hand off recent work, or check what they actually own versus what AI carried. Skip decisions already handed over — a build-time probe counts as handed over.
Resolves technical HOW decisions when the intent is clear but the technical path is not — architecture, technology selection, design patterns.
| name | moo-authoring |
| description | Use when adding or changing a skill, fragment, hook, or runtime file in this repo — including deciding which unit a new capability should be. |
Doctrine for authoring moo's own surfaces.
A skill's frontmatter carries name and description; read any shipped SKILL.md for the current key set. The description is one line, and Claude Code caps it at 1024 characters.
Version lives in plugin.json only (DRY). The official Claude Code spec does not allow version in SKILL.md frontmatter.
hope/skills/card.md sets for cards.Phrase design decisions as "X over Y: reason".
Unit choice — behavior inlines at build, data references at runtime; a skill's firing is probabilistic, so behavior that must run every time is a hook:
| The new thing is... | Unit |
|---|---|
| A contract that must read identically in ≥2 skills | Fragment (<plugin>/skills/*.md, added to sync --files) |
| Data selected per use (catalog, profile, corpus) | Runtime file |
| A trigger + procedure that stands alone | Skill |
| A trigger only the human perceives | Skill with disable-model-invocation: true |
| Behavior that must run every time, deterministically | Hook |
| An unproven idea | hunch skill + HYPOTHESIS.md; graduates or dies |
Composition — artifacts and priming over imports:
Hooks run beside the thread, never in front of it — they never gate.
Mode = two questions: does the foreground need the result, and now?
| Mode | When |
|---|---|
| Sync inject | Few lines of framing, computed instantly |
async — fire-and-forget | Side effect only; surfaces as a file change, never re-engages the thread |
asyncRewake — fire-and-maybe-wake | Off-thread check; exit 2 wakes Claude on a finding, exit 0 stays silent |
A hook that spawns headless claude -p copies its flag set from the two shipped hooks, where each flag is commented at the point of use (hope/hooks/judge.sh, hope/hooks/memory-write.sh), and keeps its verdict logic in one file shared with its eval harness.