sase-notify

name	sase_notify
description	Inspect SASE notifications and notification inbox entries. Use when the user asks about SASE notifications, the notification inbox, axe notifications/errors, plan or question notifications, or "the notification I just got".

Before doing anything else, run this command to record that you are using this skill:

sase skill use sase_notify --reason "<one-line reason for using this skill>"

Quick reference for inspecting SASE notifications. This skill is read-only: do not dismiss, mute, snooze, mark read, or otherwise mutate notifications unless the user explicitly asks and a SASE CLI command exists for that action.

Primary command

sase notify list -j -l 20

This prints a stable-shape JSON array of recent notifications, newest first. Each row has: id, timestamp, age, sender, priority, notes, files, action, action_data, read, dismissed, silent, muted, and snooze_until.

Prefer JSON for discovery and filtering. Useful examples:

• sase notify list -j -l 20 --sender axe — recent axe notifications.
• sase notify list -j -l 20 --unread — unread notifications only.
• sase notify list -j -l 20 --sender axe --unread — unread axe notifications.
• sase notify list -j -l 20 -q '<text>' — search notification ids, senders, notes, files, actions, and action data.
• sase notify list -j -l 20 --all — include dismissed notifications.

If the list is empty, say "no matching notifications found" plainly. Do not fabricate notification context.

Exact inspection

After choosing a notification id, inspect it directly:

sase notify show --id <notification_id>

The default output is markdown and includes the notification id, timestamp, age, sender, priority, notes, attached files, action, action data, and state flags. Use JSON when automation needs the exact projection:

sase notify show --id <notification_id> -f json

If show reports that the notification was not found, say so plainly instead of falling back to an unrelated notification.

Command-backed gate notifications

PlanApproval, EpicApproval, UserQuestion, LaunchApproval, and CustomGate are typed projections of command-backed interaction gates. Their action_data includes stable request identifiers and owned paths; rich definitions and reviewed content stay in the neutral interaction_requests/<kind>/<request-id>/request.json bundle. Inspect those identifiers when they help answer the user's question, but do not write response.json, invoke bundle commands, or mutate a pending action by hand. ACE, mobile, Telegram, and the typed CLI commands resolve and execute the same validated gate. Use /sase_gate to author a custom gate that proposes commands for user confirmation.

sase gate create is the creation API for a registered gate specification on stdin; sase gate wait waits for its terminal result. Ordinary raw notifications use sase notify create; their JSON may include "silent": true, which keeps the audit row while hiding it from live delivery surfaces. Raw creation cannot mint a privileged typed gate action. This skill remains read-only unless the user explicitly asks for a supported notification mutation or creation.

Axe digest notifications

For axe digest or error notifications, the actionable details are usually in an attached digest file. After identifying the notification:

1. Check files for an error digest path.
2. Check action_data.error_report_path if files is empty or ambiguous.
3. Read the attached file before summarizing the actual error details.

Distinguish notification notes from attached file contents. The notification notes are the inbox summary; the digest file is the detailed report.

How to summarize

• Cite the notification id, sender, and timestamp you used.
• Mention whether the notification is unread, dismissed, silent, muted, snoozed, or priority when that state affects the answer.
• For axe digests, cite the attached file path you read and summarize only what it contains.
• Keep direct quotes short; prefer concise paraphrases unless exact wording matters.
• Report uncertainty and missing data directly. Do not infer hidden notification contents from sender names or actions.