hermes-labs-ai

hermes-labs-ai/little-canary

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.

4 good first / help-wanted issues · Python · last activity Sep 29, 2026

36 stars 7 forks 36 watchers Python Apache License 2.0
agent-security agent-skills ai-agents ai-safety canary claude-code-plugin defense-in-depth detection developer-tools gemini-cli-extension hermes-labs input-validation jailbreak-detection llm-security mcp preflight prompt-injection prompt-injection-detection python
4 Open Issues Need Help Last updated: Sep 29, 2026

Open Issues Need Help

View All on GitHub
documentation help wanted
hermes-labs-ai/little-canary
36

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.

Python
#agent-security#agent-skills#ai-agents#ai-safety#canary#claude-code-plugin#defense-in-depth#detection#developer-tools#gemini-cli-extension#hermes-labs#input-validation#jailbreak-detection#llm-security#mcp#preflight#prompt-injection#prompt-injection-detection#python
hermes-labs-ai/little-canary
36

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.

Python
#agent-security#agent-skills#ai-agents#ai-safety#canary#claude-code-plugin#defense-in-depth#detection#developer-tools#gemini-cli-extension#hermes-labs#input-validation#jailbreak-detection#llm-security#mcp#preflight#prompt-injection#prompt-injection-detection#python
hermes-labs-ai/little-canary
36

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.

Python
#agent-security#agent-skills#ai-agents#ai-safety#canary#claude-code-plugin#defense-in-depth#detection#developer-tools#gemini-cli-extension#hermes-labs#input-validation#jailbreak-detection#llm-security#mcp#preflight#prompt-injection#prompt-injection-detection#python
enhancement good first issue
hermes-labs-ai/little-canary
36

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.

Python
#agent-security#agent-skills#ai-agents#ai-safety#canary#claude-code-plugin#defense-in-depth#detection#developer-tools#gemini-cli-extension#hermes-labs#input-validation#jailbreak-detection#llm-security#mcp#preflight#prompt-injection#prompt-injection-detection#python