coyote-strategy

Installation
SKILL.md

🐺 COYOTE v1.0.0 β€” Regime Classifier / Meta-Router

The agent that asks "what kind of market are we in?" before anything else. Coyote watches macro conditions and classifies the market into TREND_UP / TREND_DOWN / CHOP. It takes a single positional expression of its regime view β€” and publishes that view on every tick.

Why this strategy exists

Every Senpi agent currently re-implements its own version of "should I be active right now?" Vulture has a HEAVY_FLOW gate. Wolverine has a multi-timeframe alignment check. Tortoise just buys regardless. None of them share a centralized view of the macro regime β€” they each make local calls.

Coyote is the start of fixing that. Its classification is emitted in every signal data block (and in every tick output, including when no trade fires), so:

  • Today: operators can read Coyote's regime view directly to inform manual decisions about which agents to enable/pause.
  • Soon (runtime work pending): other agents will be able to subscribe to Coyote's regime channel as a gating input, so e.g. Tortoise auto-pauses in TREND_DOWN and Stag auto-enables in TREND_UP.

The "meta-router" framing is aspirational for v1 β€” Coyote also takes its own regime-positional trade so it has skin in its own calls.

CRITICAL RULES

RULE 1: Three regimes, two trades

Coyote classifies into exactly three states:

Installs
5
GitHub Stars
99
First Seen
Jun 7, 2026
coyote-strategy β€” senpi-ai/senpi-skills