Menú
Capability discovery — the canonical answer to 'what can you do?' and 'what do I say to ...?'
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
Close skill — draft resolution comment, sync backends, redirect to next task
Start skill — begin task with optional backend sync
End-of-day wrap-up skill — daily summary and status update
Morning brief skill — prioritized daily summary
Reconcile skill — detect wiki/backend drift and let the User resolve it via four moves
Triage skill — walk untriaged Asana pages, push a new ticket or mark not-needed
| name | help |
| description | Capability discovery — the canonical answer to 'what can you do?' and 'what do I say to ...?' |
| risk_class | soft |
The canonical "now what?" answer. Lists every capability the vault has, organized by what the user wants to do, with natural-language phrasings that trigger each. Zero token burn unless invoked — no proactive tips, no daily banners, no inline reminders.
Also serves as the author's reference when they forget what they've wired
up. It's literally /help's job to be the index.
help, /help, what can you do, what do I say to …,
how do I …, commands, or any equivalent phrasing./onboard./help and quote the relevant section back rather than
improvise.Render a single, scannable message. Group capabilities by what the user wants to do. Within each group, list the natural-language phrasing first and the underlying skill / CLI second — the user thinks in intent, not in verbs.
/good-morning — generate today's
daily page and a prioritized brief; sweeps backend drift via /reconcile
before gathering context./wrap-up — summarize the day, run
the linter, sweep backend drift via /reconcile after the day's local
writes, and update tomorrow's focus.Both routines invoke /reconcile as part of the twice-daily reconciliation
ritual — see Workflows at the wiki ↔ backend seam
for the full design walk-through.
/start — flips the task to in-progress, logs
it on today's daily page./close — marks the task done,
stamps the close date, logs the activity./link —
cross-link two tickets in the same backend./transition — sync the wiki and
backend statuses./asap — file an ASAP item that persists
until resolved./idea — capture an idea for later./remind — date-keyed reminder./meeting-note — file a structured meeting note./ingest-writing —
paste any raw material (drafts, emails, articles, project docs). The
skill extracts what's useful and routes it into wiki/voice.md,
about.md, vocabulary.md, or preferences.md. Append-only./query — natural-language search of the work history./lint — workspace health report./connect — wire up an external system. Surfaces
four options inline (official CLI / pp-CLI / MCP / hand-written) and
lets you pick./reconcile —
refresh active tasks, surface drift between wiki and each backend, and
walk the User through resolution one ticket at a time./push — draft, preview-confirm, create
the external ticket, back-link, audit. Gated by confirm./comment — post a comment with a
preview-confirm gate./new-ticket — originate a brand-new ticket on
the intake backend, optionally mirror downstream./release — log a repo release and
optionally finalize the included tickets across backends.Every setting has a natural-language phrasing. Translating user intent into
the right rubber-ducky settings ... call is part of normal conversation;
this section tells the user what knobs exist.
| Setting key | What it controls | Say this to change it |
|---|---|---|
ingest.auto_on_wrap_up | At /wrap-up, offer to ingest today's drafts as voice samples. | "start / stop ingesting at wrap-up" |
ingest.auto_on_onboard | At /onboard, invite the user to paste raw material. | "ask / don't ask me to paste at onboard" |
ingest.kinds | Which extraction categories /ingest-writing produces. | "only extract voice and vocabulary" |
confirm.<transport>.<verb> | Per-action confirm policy — auto (skip prompt) or preview (show + ask). Unlisted actions default to preview. | "always auto-confirm Jira comments" |
Operators can read or write any setting directly:
rubber-ducky settings get <path> # read a value
rubber-ducky settings set <path> <value> # write a value (type-validated)
rubber-ducky settings enable <path> # shorthand for `set <path> true`
rubber-ducky settings disable <path> # shorthand for `set <path> false`
Every change is audit-logged to wiki/log.md.
The vault splits actions into two risk classes:
risk_class: soft — reads, drafts, writes landing in the local
vault. Auto-executes.risk_class: hard — sends, posts, creates, updates touching external
systems. Gated by rubber-ducky confirm. The skill prepares a structured
preview, calls confirm request --action <descriptor> --preview <text>
(the CLI shows the preview and prompts the user unless
settings.confirm.<descriptor> is auto), and only proceeds after
confirm consume validates the token.Per-action overrides under confirm.* are the power-user knob — say
"always auto-confirm Jira comments" or "always preview Slack DMs."
Most users run claude from a normal terminal or Claude Code Desktop. For
a Chase Hannigan-style visual chrome you can run rubber-ducky inside
Obsidian:
claude and
the working directory to the vault root.The vault stays browsable in Obsidian's editor while claude runs in the
in-Obsidian terminal pane. Behavior is identical to running from a
standalone terminal.
The end-user surface is conversational. These exist for operators debugging or scripting:
rubber-ducky init <dir> — bootstrap a new vault.rubber-ducky doctor / doctor lint — workspace health checks.rubber-ducky status — show vault info.rubber-ducky update — refresh bundled skills.