First Seen
Feb 18, 2026
Last Scanned
Feb 20, 2026
Findings
4
Score
77/100
Findings (4)
Detects instructions to hide actions from the user
Don't share tokens** - Each userRemove directives that attempt to change the agent's output format, suppress safety warnings, or alter response structure in ways that bypass safety controls.
Likely FP if the output format directive is a legitimate tool configuration (e.g., return results as JSON) that does not suppress safety features.
Detects skills that both read sensitive credential files and send data to external services
load = json.loads(os.env + Send Results via APIBlock access to git credentials, SSH keys, and repository tokens. If git operations are needed, use scoped deploy keys and restrict the tool to specific repositories.
Likely FP if the match is documentation about git configuration (e.g., setting up git credentials helper) rather than code that reads and transmits them.
Detects patterns where external API responses are used directly without validation or sanitization
Api
from datasets import Data + automatically available if passValidate and sanitize all data received from external APIs before using it in tool operations or agent prompts. Implement schema validation and treat API responses as untrusted input.
Likely FP if the match is a truncated table cell or documentation fragment that mentions API responses in a descriptive context, not actual unvalidated data processing.
Detects references to raw.githubusercontent.com on mutable branches like main/master
raw.githubusercontent.com/huggingface/trl/main/Replace GitHub raw.githubusercontent.com references with pinned commit SHAs instead of branch names (e.g., /commit-sha/file instead of /main/file). Branch references are mutable.
Likely FP if the raw GitHub URL points to a versioned release tag in a well-known repository, though even tags are technically mutable.