little-canary
SkillGitHubActifPython
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.
Mesuré: 24 oct. 2026 · 43 observations
Domaines & Usages
- Domaines
- Agents, skills et MCP
- Usages
- Compétences d’agents
Installation
/plugin marketplace add hermes-labs-ai/little-canaryÀ vérifier à la source
Compatibilité et prérequis sont à vérifier dans la documentation officielle.
https://github.com/hermes-labs-ai/little-canaryClassifications
- Nature
- Outil
Historique des mesures
Les relevés comparables sont regroupés par signal et unité.
Étoiles
40étoiles · 24 oct. 2026
+12 (+42,9 %)depuis le 11 sept. 2026
Valeurs exactes · 43 relevés
| Mesuré | étoiles | Écart |
|---|---|---|
| 40 | 0 | |
| 40 | +1 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | 0 | |
| 39 | +1 | |
| 38 | 0 | |
| 38 | 0 | |
| 38 | 0 | |
| 38 | 0 | |
| 38 | 0 | |
| 38 | 0 | |
| 38 | +2 | |
| 36 | 0 | |
| 36 | 0 | |
| 36 | +2 | |
| 34 | +1 | |
| 33 | +1 | |
| 32 | 0 | |
| 32 | 0 | |
| 32 | +2 | |
| 30 | 0 | |
| 30 | 0 | |
| 30 | 0 | |
| 30 | 0 | |
| 30 | 0 | |
| 30 | 0 | |
| 30 | +1 | |
| 29 | +1 | |
| 28 | 0 | |
| 28 | · |