UnplugAI/Unplug

ScanResult.degraded is never set 9 days ago

bug help wanted mentored

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

The composite action is tag-pinned and dependabot has never updated it 14 days ago

good first issue needs-triage

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

docs-requirements.txt has no dependabot watcher 16 days ago

good first issue needs-triage

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Cancel superseded CI runs with a concurrency group 23 days ago

good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Add a security-model reading order to GETTING_STARTED 24 days ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Five documented claims no longer match the code 25 days ago

documentation good first issue needs-triage

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Flat UNPLUG_* env vars are dropped when the config file has a [guard] table 25 days ago

bug help wanted needs-triage

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Taint tracking never fires for MCP or CamelCase tool names 25 days ago

bug help wanted needs-triage

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Invalid configured model path silently falls back to the cache 25 days ago

bug help wanted

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

_fail_closed results carry no stage or latency information 27 days ago

good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Two docs still show Guard.with_tiny() 30 days ago

documentation good first issue needs-triage

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Benign benchmark corpus has no hard negatives about 1 month ago

good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Evasion classes with no matching span produce no finding at all about 1 month ago

bug help wanted

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Vendor mermaid instead of loading it from a CDN about 1 month ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Add social cards and a favicon about 1 month ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Search does not match underscored identifiers about 1 month ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Publish the community docs on the site about 1 month ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Root README and two scripts still document Guard.with_tiny() about 1 month ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Duplicate warning blocks in the ML load path about 1 month ago

good first issue claimed

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Add a pre-commit config about 1 month ago

good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Document that scan_output_request ignores request.scanners about 1 month ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

fail_closed config field is read nowhere about 1 month ago

documentation good first issue

UnplugAI/Unplug

6

Every way an LLM can break one SDK to stop it. Injection. Hallucination. Leakage. Handled.

Python

#ai-safety#llm#prompt-injection#python#sdk#security

Open Issues Need Help