← All creators

kylus

User

Make a long-lived agent portable: swap the model provider underneath it, swap the chat channel in front of it, and gate memory writes behind human approval.

6 indexed · 0 Featured · 0 stars · avg score 69
Prolific

Categories

Indexed Skills (6)

AI & Automation Listed

approve-proposal

Owner-only — review a pending proposal, optionally edit, then merge into `memory/<section>.md`. Also handles rejection. Triggers: "approve", "merge proposal", "拒絕提案", "/approve <file>", "/reject <file>".

0 Updated 3 weeks ago
kylus
AI & Automation Listed

ingest-source

Owner-only — pull a single GitHub/GitLab item, Google Doc, local file, or (if the channel is enabled) Slack thread into `sources/`, then stage a memory-update proposal that the owner can refine and commit. Triggers: "ingest <url>", "吸收這個", "/ingest", URL paste from owner.

0 Updated 3 weeks ago
kylus
AI & Automation Listed

propose-memory-update

Contributor flow — write a proposal to `pending/` for the owner to review. Contributors cannot edit `memory/*.md` directly. Use when the user requests a memory change but is not the topic owner (role=contributor). Triggers: "propose update", "建議補上", "我想加進記憶", "/propose-memory-update".

0 Updated 3 weeks ago
kylus
AI & Automation Listed

support-thread

Analyze a support/incident chat thread and emit a structured triage: problem, environment, what's been tried, current hypothesis, requester expectations, next step, owner, risks, sources. Also classifies any commitment-shaped statements (confirmed / proposed / internal / unclear). Triggers: "整理 support thread", "客服", "/support".

0 Updated 3 weeks ago
kylus
AI & Automation Listed

meeting-notes

Turn a meeting transcript or chat-thread-of-a-meeting into structured notes: background, key discussion points, decisions, todos, open questions, risks, understanding changes, and a short summary pastable back to the channel. Triggers: "整理會議", "meeting notes", "transcript", "/meeting".

0 Updated 3 weeks ago
kylus
AI & Automation Listed

read-memory

Dump the current state of this topic's section-versioned memory (background, current_understanding, decisions, open_questions, commitments, people). Use at the start of every Q&A turn so the agent answers from memory, not training data. Triggers: "what do we know", "current state", "read memory", "目前理解", "現在進度", "/read-memory".

0 Updated 3 weeks ago
kylus

Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.