little-canary

SkillGitHubActivePython

Detects prompt injection by its effect on a sacrificial canary model, not just pattern matching: untrusted input hits a powerless model first, a behavioral check reads the residue, and it returns block, flag, or pass before your primary model acts. Inbound preflight sensor, not a guarantee.

Description provenance: Original source

Open canonical source

Measured: Oct 24, 2026 · 43 observations

Domains & Uses

Domains: Agents, skills and MCP
Uses: Agent skills

Installation

/plugin marketplace add hermes-labs-ai/little-canary

Check at the source

Compatibility and prerequisites must be checked in the official documentation.

https://github.com/hermes-labs-ai/little-canary

Classifications

Nature: Tool

Measurement history

Comparable readings are grouped by signal and unit.

Stars

40stars · Oct 24, 2026

+1 (+2.6 %)since Oct 17, 2026

Exact values · 8 readings

Measured	stars	Change
Oct 24, 2026	40	0
Oct 23, 2026	40	+1
Oct 22, 2026	39	0
Oct 21, 2026	39	0
Oct 20, 2026	39	0
Oct 19, 2026	39	0
Oct 18, 2026	39	0
Oct 17, 2026	39	·