Skrya is a briefing skill pack for AI agents. The user describes what they want to track; the agent turns that into a durable topic with confirmed scope, confirmed sources, recurring delivery, daily digests, follow-up analysis, and long-term feedback memory.
User: Track important AI browser developments for me every day.
Skrya turns it into:
1. topic scope: AI browsers, browser agents, search/browser distribution shifts
2. source policy: official releases, major tech media, product updates, social signals
3. recurring briefing: delivered back to the original conversation or channel
4. daily digest: uniform format, compact sources, follow-up commands
5. follow-up threads: continuing stories stay connected instead of resetting every day
# 2026-05-08 | AI Browsers | Daily Briefing
┌─ **【Brief 1】OpenAI browser rumors matter because the browser can become the agent runtime**
│ Judgment: the durable shift is not "another browser"; it is identity, page state, and tool use moving into one workflow.
│
│ Sources: official blog / major tech media / product pages
└
┌─ **【Brief 2】Perplexity, Arc, and adjacent products keep blending search, browsing, and agent actions**
│ Judgment: AI browsers are a distribution change for search and task execution, not a single feature category.
│
│ Sources: product updates / media reports / user feedback
└
---
## System
- Reply `dig: 1` for deeper analysis of item 1.
- Reply `track: 2 AI browser product line` to keep item 2 as a continuing thread.
- Reply `fewer low-quality rumors` to update future briefing preferences.See more demos: AI browser daily briefing, new energy thread, and research agent watchlist.
Send this to your agent:
Install Skrya from https://github.com/Arcadia822/skrya. After installation, send me a daily briefing about important AI browser developments.
The agent should read the repository installation instructions and handle the setup. You should not need to copy shell commands manually. After installation, it should clarify what you actually want, then confirm sources and delivery. It should not dump search results immediately. We are all grateful for this restraint.
Manual install:
git clone https://github.com/Arcadia822/skrya.git
cd skrya
python3 -m pip install -e .
./setup --host autoTurn a standing area of interest into a reusable daily briefing. Each item has source references, judgment, and follow-up actions.
Skrya confirms sources before declaring a topic ready. It distinguishes automatically connected sources, runtime retrieval sources, and sources that cannot be connected yet.
Skrya looks for useful local skills and uses them to widen source coverage. For example, it can work with agent-reach for public web and social retrieval, or wechat-article-search for WeChat Official Account articles. Skrya does not focus on bypassing login walls or access restrictions; it organizes available sources into reusable briefing workflows.
The same developing story should not become unrelated headlines every day. Skrya links related items into a continuing thread so the context stays visible.
When you say "show fewer items like this", "keep tracking this", or "expand item 3", Skrya turns that into durable preferences instead of treating it as disposable chat.
In hosts with channels or conversations, a briefing created in one channel stays bound to that channel by default. Resends and empty-briefing diagnosis do not pull in another channel's topic.
Skrya does not need an install-time language setting. Language belongs to the topic:
- New topics default to the language used when the topic is created.
- You can explicitly ask for a topic to output in Chinese or English.
- The project currently supports Chinese and English output.
- If you give feedback in another language later, the agent should answer in your feedback language without changing the topic's briefing language unless you ask.
Skrya runs fully locally. It has no telemetry, cloud sync, or hosted backend, and it does not proactively upload your topic config, briefing history, feedback preferences, or local data. Other networked skills you ask the agent to combine with Skrya, such as web retrieval, browser tools, or WeChat article search, follow their own external-access rules.
Skrya supports three uninstall modes:
| Mode | Removes | Keeps |
|---|---|---|
skills-keep-data |
Installed skills | Topic config, history, data-root config |
data-keep-skills |
Skrya data and data-root config | Installed skills |
complete |
Skills, data, and Skrya routing notes in global instructions | Unrelated user instructions |
Tell your agent:
Uninstall Skrya. First explain the three uninstall modes, then wait for my confirmation before acting.
Complete uninstall removes only marked Skrya instruction blocks such as SKRYA-ROUTING-NOTE. It should not wipe unrelated AGENTS.md, CLAUDE.md, TOOLS.md, or tools.md content.
| Doc | Covers |
|---|---|
| Install | Agent-facing install, uninstall, and upgrade steps |
| User journeys | How agents should behave for nontechnical users |
| Threads | Continuing event timelines |
| Delivery bindings | How topic delivery stays isolated to the creating channel and user context |
| External retrieval | Runtime retrieval normalization |
| Domain model | Topic, request, source, thread, channel, and related entities |
| Contributing | Development, tests, code layout, and technical design |