routine-from-chat

routine-from-chat is a skill for Claude Code, Codex from rome-os/rome. It costs 124 tokens per session (3,917 once invoked), scanned A, original, MIT.

A skill that turns a plain-language request for automation into an event-triggered, scheduled, or manually run routine.

In plain words
What is it for?
Use it to create requests such as notifying you when an email arrives, reminding you every Friday, or saving a playbook to run by hand.
Why use it?
It removes the need to know trigger names, schedules, time zones, filters, or JSON when describing an automation.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Good fit Use it to create requests such as notifying you when an email arrives, reminding you every Friday, or saving a playbook to run by hand.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/rome-os/rome/routine-from-chat
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add rome-os/rome --skill routine-from-chat
Clone the repo
git clone --depth 1 https://github.com/rome-os/rome

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for routine-from-chat

README.md
[![agentmods](https://agentmods.dev/badge/skills/rome-os/rome/routine-from-chat.svg)](https://agentmods.dev/skills/rome-os/rome/routine-from-chat)
Your own site
<a href="https://agentmods.dev/skills/rome-os/rome/routine-from-chat"><img src="https://agentmods.dev/badge/skills/rome-os/rome/routine-from-chat.svg" alt="Measured on agentmods" height="20"></a>
Per session 124 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,917 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00124 $0.03917
Opus 5 $0.00062 $0.01959
Sonnet 5 $0.00025 $0.00783
Haiku 4.5 $0.00012 $0.00392

Measured 8d ago against content hash 45e5be846764, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

routine-from-chat scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

rome_apps/assistant/skills/routine-from-chat/SKILL.md · 270 lines

How it starts

The opening of the file, as written. The whole thing — 270 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Create a routine from chat

When the guardian describes something they want to happen automatically, turn that intent into a routine — without making them learn event names, RRULEs, timezones, filters, or JSON. They speak in their language; you map it to a concrete trigger, ask only for what you genuinely can't infer, then show a confirm card. You do not create the routine yourself — the card creates it when they turn it on.

A routine fires one of three ways. Decide which the request is:

  • Event — "when X happens": an email arrives, a payment lands, a form is submitted. Reactive, no fixed time.
  • Schedule — "at a time / on a cadence": "every Friday at 9am", "tomorrow at noon", "the 1st of each month", "every morning".
  • Manual — "save this so I can run it whenever": a playbook with no automatic trigger. the guardian runs it by hand from the Routines page's "Run now" button.

If it's ambiguous ("remind me about the standup"), ask one short question to disambiguate before going further. In particular, if it's unclear whether they want it on a schedule or to run it by hand, ask.

The flow

  1. Classify the request as event, schedule, or manual (above).
  2. Resolve the trigger.
    • Manual: nothing to resolve — there is no trigger to watch or time to compute. Skip straight to choosing the Then (the action it runs) and propose with kind: "manual". No eventName, tzid, rrule, or date.
    • Event from a third-party app (Gmail, GitHub, Slack, Stripe, …): go through the connector chain — connector_event_searchconnector_event_schemaconfirm with the guardianconnector_event_subscribe. It finds the trigger, gets a yes before arming it upstream, and hands you the exact eventName to bind. See "Event routines from a connector".
    • Event already flowing on the bus (a Rome app's own event, or a connector event you set up before): call search_event_catalog with keywords from the request, then use the matching eventType as eventName. Never invent an event name.
    • Schedule: work out the cadence, the time of day, and the timezone from what they said; translate to an RRULE / one-off date behind the scenes.
  3. Fill the gaps with ask_question. Underspecified requests are the norm ("important", "my landlord", "in the morning"). Ask one short, recognition-style question per gap: a single-type question with concrete options when you can name them (candidate senders; "8am / 9am / noon"), or a text question when the answer is open-ended. Never demand a raw value the user would have to look up, and never ask them for an RRULE or an IANA timezone — ask "what time?" and "roughly where are you?" and convert it yourself.
  4. Propose with propose_routine. Pass kind: "event", kind: "schedule", or kind: "manual" plus the human summary and the machine spec. A confirm card renders; the guardian turns it on. For manual, set watchLabel to something like "Run on demand" and omit all trigger fields.
  5. Close. After propose_routine, reply with one short line confirming what you drafted, then stop. Do not call any create action — the card does that.

Read the full file on GitHub · 270 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 8d ago First seen · 270 lines · 124 tokens per session scan A 45e5be846764

Subscribe to this mod's changes

routine-from-chat is a skill published in the GitHub repository rome-os/rome (482 stars, last pushed today), licensed MIT. It adds 124 tokens to every session and 3,917 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.