AI coding agent skills scoped to polily user lifecycle support — onboarding, daily analysis follow-ups, custom strategy authoring, troubleshooting. Works in Claude Code, Hermes, and OpenClaw. Install the pack once; the skill auto-activates in your agent session when you reference polily, no re-explaining each time.
Without the plugin, asking your agent about your polily install means starting from scratch every chat:
"What's the schema of polily's
analysestable? It's a Polymarket monitoring tool I run locally; the DB is at..."
With the plugin loaded, your agent already knows polily — DB schema, daemon mechanics, file paths, and how polily's analytical agent thinks. You ask, it answers, grounded in your actual polily.db state:
"Why did polily say BTC market edge was thin in yesterday's analysis?" "Show me my biggest open position and how it's tracking against the entry thesis." "What's my realized PnL this week, by event?" "I want to write a custom analysis strategy — walk me through the active_strategy mechanism."
Your agent consults the skill's reference knowledge, runs SQL against your local polily.db, fetches polily's analytical methodology at runtime (your custom strategy if you've authored one, polily's official default.md otherwise), and answers in polily's own framing — consistent with whatever you just read in the TUI.
Pick your agent CLI:
All three install paths give the same polily skill — same SKILL.md, same DB schema reference, same methodology lookup. The differences are only in how each harness discovers and loads plugins.
In a Claude Code session:
/plugin marketplace add ShiyuCheng2018/polily-plugin
Once you see "Successfully added marketplace: polily-plugin", run:
/plugin install polily@polily-plugin
Then activate it without restarting the session:
/reload-plugins
The skill auto-activates next time you mention polily in chat.
→ See docs/install-claude-code.md for marketplace URL variants, modal fallback, troubleshooting, and uninstall.
Two install paths:
A. git clone + config (recommended for development tracking)
git clone https://github.com/ShiyuCheng2018/polily-plugin ~/polily-pluginAdd to ~/.hermes/config.yaml:
skills:
external_dirs:
- ~/polily-plugin/skillsRestart Hermes. Verify with: use skill tool to list skills (should show polily).
B. Hermes tap (native, no config editing)
Hermes installs skills directly from a GitHub "tap". This repo already has the required skills/polily/SKILL.md layout — no submission step:
hermes skills tap add ShiyuCheng2018/polily-plugin
hermes skills install ShiyuCheng2018/polily-plugin/skills/polily
→ See docs/install-hermes.md for troubleshooting and uninstall.
A. ClawHub install (pending submission, tracked for v0.2.x)
clawhub install polilyB. git clone + extraDirs
git clone https://github.com/ShiyuCheng2018/polily-plugin ~/polily-pluginAdd to your OpenClaw config:
skills:
load:
extraDirs:
- ~/polily-plugin/skillsRestart OpenClaw. Verify the skill is discovered.
→ See docs/install-openclaw.md for troubleshooting and uninstall.
Requires polily installed locally (pipx install polily) for most skill features — the skill falls back to fetching from GitHub when polily isn't installed, but local-install experience is faster and more accurate.
When a new version ships (releases page):
/plugin marketplace update polily-plugin
/plugin update polily@polily-plugin
Or update everything at once:
/plugin marketplace update
/plugin update --all
There's a known Claude Code issue (anthropics/claude-code#21995) where skipping the first step makes /plugin update report "already at the latest version" even when a new release exists. The two-step is the reliable path.
→ See docs/install-claude-code.md for full troubleshooting.
- tap install: re-run
hermes skills install ShiyuCheng2018/polily-plugin/skills/polilyto pull the latest (the tap tracks this repo) - git clone install:
cd ~/polily-plugin && git pull
- ClawHub install (when available):
clawhub update polily - git clone install:
cd ~/polily-plugin && git pull
| Skill | What it does | Source |
|---|---|---|
skills/polily/ |
Reference for polily users chat-querying their install — DB schema, daemon mechanics, file paths, runtime methodology lookup (consults user's active strategy or polily's default.md at chat time) |
polily/agents/skill_sources/core/ |
As polily evolves and new lifecycle-stage skills land, they get added here. See CLAUDE.md for the inclusion bar and contribution flow.
The polily skill activates when you mention polily by name in chat, or reference polily-specific concepts (structure_score, polily.db, next_check_at, etc.) — in any supported harness (Claude Code, Hermes, OpenClaw). It does not activate on generic Polymarket questions unrelated to polily — that boundary is encoded in the skill's YAML description so the plugin stays scoped.
Examples that activate:
- "Why did polily flag this event as Tier C?"
- "Show me my polily positions"
- "polily structure_score 怎么算的？"
Examples that don't:
- "What is Polymarket?" (no polily reference; generic platform question)
- "How does Kalshi compare to Polymarket?" (no polily reference; cross-platform comparison)
Each skill file carries a GENERATED FILE — DO NOT EDIT header naming its source repo. Hand-edits to generated skills are overwritten on the next regeneration. To change a generated skill's content, edit the source files in the source repo and run its generator:
cd path/to/polily
python scripts/generate_skills.py --plugin-repo path/to/polily-pluginThe generator writes to both polily's internal copy AND this repo's artifact. See CLAUDE.md for full rules.
This pack releases independently of polily. Each release bumps the version across all present harness manifests (use ./scripts/bump-version.sh X.Y.Z) and gets a corresponding git tag. See CHANGELOG.md and the releases page for history.