How do I install the Alert skill?

Run npx skillmds@latest add jasoncarreira/alert in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Alert safe to use?

SkillMD's automated safety review verdict for this skill is PASS. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Alert?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Alert free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

jasoncarreira (@jasoncarreira) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Alert by jasoncarreira — Agent Skill

Name: Alert
Rating: 4.2 (5 reviews)
Author: jasoncarreira

Alert

Use when a high-priority signal needs to reach the operator but doesn't fit the current conversation (off-hours wake, urgent error needing eyes, scheduled-tick discovery worth surfacing now). Routes via MIMIR_OPERATOR_ALERT_CHANNEL.

jasoncarreira Updated Aug 19, 2026 6 repo stars

File contents

Operator alert channel

If MIMIR_OPERATOR_ALERT_CHANNEL is configured, the system prompt's Operator config section will name a channel id (e.g. dm-slack-D05ABC or dm-discord-NNN — a DM conversation id, not a user id). Use it for high-priority signals to the operator that don't fit the current conversation. If the line is absent, the feature is off — fall back to your usual options (heartbeat backlog, the channel you're already in).

Use it for

Critical errors you can't recover from on your own
Urgent findings during a heartbeat the operator should know about now
Dispatch failures or systemic issues that need human attention
Time-sensitive escalations where the cost of waiting is high

Don't use it for

Routine updates ("I did a thing today")
Status reports or check-ins
Low-priority observations (those go in state/heartbeat-backlog.md)
Anything that can wait for the next time the operator messages you

How

The alert channel is a normal channel id — the registered bridge dispatches by prefix. So just call send_message against it like any other channel:

send_message(channel_id="dm-slack-D05ABC", text="...")

Make the message specific and self-contained. The operator may be reading it out of context, so:

Lead with what happened ("Discord dispatch failing for the past 4 hours")
Include the key data ("DiscordError: Forbidden, ts: 2026-05-01T14:02Z")
Say what you've already tried, if anything
End with what you need from them, if anything

Volume calibration

If you find yourself using this channel more than once a day, you're probably misusing it — re-read the "Don't use it for" list. The signal loses meaning when it goes off too often.

When the channel is misconfigured

If MIMIR_OPERATOR_ALERT_CHANNEL points at a channel id no bridge recognizes, send_message returns is_error: true and logs a send_message_unknown_channel event in events.jsonl — but the operator never sees the alert. There's no auto-fallback. Treat silent loss as the failure mode and verify the channel id at onboarding (the operator can send_message to it themselves to confirm). If you observe send_message_unknown_channel events in the algedonic feedback block, raise it as a real issue — the alert channel itself has gone dark.

---name: alertdescription: Use when a high-priority signal needs to reach the operator but doesn't fit the current conversation (off-hours wake, urgent error needing eyes, scheduled-tick discovery worth surfacing now). Routes via MIMIR_OPERATOR_ALERT_CHANNEL.---# Operator alert channelIf `MIMIR_OPERATOR_ALERT_CHANNEL` is configured, the system prompt'sOperator config section will name a channel id (e.g. `dm-slack-D05ABC` or`dm-discord-NNN` — a DM *conversation* id, not a user id). Use it for high-priority signals to the operator that*don't fit the current conversation*. If the line is absent, the featureis off — fall back to your usual options (heartbeat backlog, the channelyou're already in).## Use it for- Critical errors you can't recover from on your own- Urgent findings during a heartbeat the operator should know about now- Dispatch failures or systemic issues that need human attention- Time-sensitive escalations where the cost of waiting is high## Don't use it for- Routine updates ("I did a thing today")- Status reports or check-ins- Low-priority observations (those go in `state/heartbeat-backlog.md`)- Anything that can wait for the next time the operator messages you## HowThe alert channel is a normal channel id — the registered bridgedispatches by prefix. So just call `send_message` against it like anyother channel:```send_message(channel_id="dm-slack-D05ABC", text="...")```Make the message specific and self-contained. The operator may bereading it out of context, so:- Lead with what happened ("Discord dispatch failing for the past 4 hours")- Include the key data ("DiscordError: Forbidden, ts: 2026-05-01T14:02Z")- Say what you've already tried, if anything- End with what you need from them, if anything## Volume calibrationIf you find yourself using this channel more than once a day, you'reprobably misusing it — re-read the "Don't use it for" list. The signalloses meaning when it goes off too often.## When the channel is misconfiguredIf `MIMIR_OPERATOR_ALERT_CHANNEL` points at a channel id no bridgerecognizes, `send_message` returns `is_error: true` and logs a`send_message_unknown_channel` event in `events.jsonl` — but theoperator never sees the alert. There's no auto-fallback. Treatsilent loss as the failure mode and verify the channel id atonboarding (the operator can `send_message` to it themselves toconfirm). If you observe `send_message_unknown_channel` events inthe algedonic feedback block, raise it as a real issue — the alertchannel itself has gone dark.

Frequently asked questions