メニュー
Systematic workflow for checking and responding to notifications. Use at session start and periodically. Covers prioritization, tone guidelines, and recording interactions.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Write and read W3C Web Annotations on ATProtocol using the at.margin.annotation lexicon. Use when annotating URLs, writing reading notes on web pages, or building a public research trail. Supports single writes, batch writes via applyWrites, and reading annotations from any ATProtocol user. Triggers on annotate, annotation, margin, reading notes, or web annotation.
Manage Semble collections and cards via ATProto CLI. Use when creating, linking, or querying Semble records - collections, cards, collectionLinks. Handles network.cosmik.* lexicons with proper field validation.
Full interaction with X (Twitter) - post, read, reply, like, retweet, follow. Use when operating on X as an additional social environment alongside ATProtocol.
Guide for observing the ATProtocol network. Use when sampling the firehose, taking network pulses, monitoring feeds, or analyzing engagement patterns.
Publish public cognition records to ATProtocol. Covers 5 record types (concepts, memories, thoughts, claims, hypotheses) with schemas and a standalone script. Use when storing knowledge, recording events, writing reasoning traces, publishing structured assertions, or tracking hypotheses. Any agent on ATProtocol can use this.
Guide for managing agent memory blocks. Use when inspecting, updating, creating, auditing, or restructuring memory blocks for yourself or subagents. Covers the memory tool (self), Letta API (subagents), auditing utilization, and invoking the memory agent for major restructuring.
| name | responding-to-notifications |
| description | Systematic workflow for checking and responding to notifications. Use at session start and periodically. Covers prioritization, tone guidelines, and recording interactions. |
The Responder V2 system uses a "Queue → Draft → Send" workflow to handle notifications efficiently and prevent missed messages.
Fetch unread mentions/replies and save them to a local draft file:
uv run python -m tools.responder queue
This creates/updates drafts/queue.yaml.
Edit drafts/queue.yaml to write your replies.
response field for items you want to reply to.reply. Can be changed if needed (e.g. like not yet supported in yaml, but for now mostly for replies).Process the queue and send out drafted replies:
uv run python -m tools.responder send
response filled in.To just view notifications without queueing (debugging):
uv run python -m tools.responder check
| Priority | Source | Action |
|---|---|---|
| 1 | Cameron (@cameron.stream) | Always respond, defer to instructions |
| 2 | Comind agents (void, herald, grunk) | Read but DON'T respond (avoid loops) |
| 3 | Known agents (Magenta, Sully) | Respond thoughtfully |
| 4 | Questions about comind/ATProtocol | Respond helpfully |
| 5 | General engagement | Respond if substantive value |
DON'T:
DO:
For each notification:
After significant interactions:
from tools.cognition import write_memory
await write_memory(
'Description of what happened...',
memory_type='interaction', # or 'correction' for errors
actors=['handle1'],
tags=['relevant', 'tags']
)
Record when:
Never respond to:
These agents are part of comind. Responding creates feedback loops.
Cameron (@cameron.stream) is the administrator. Special rules: