pub struct ContentIsolationConfig {
pub enabled: bool,
pub max_content_size: usize,
pub flag_injection_patterns: bool,
pub spotlight_untrusted: bool,
pub quarantine: QuarantineConfig,
pub embedding_guard: EmbeddingGuardConfig,
pub mcp_to_acp_boundary: bool,
pub nli: NliConfig,
pub secret_masking: SecretMaskingConfig,
}Expand description
Configuration for the content isolation pipeline, nested under
[security.content_isolation] in the agent config file.
Fields§
§enabled: boolWhen false, the sanitizer is a no-op: content passes through unchanged.
max_content_size: usizeMaximum byte length of untrusted content before truncation.
flag_injection_patterns: boolWhen true, injection patterns detected in content are recorded as
flags and a warning is prepended to the spotlighting wrapper.
spotlight_untrusted: boolWhen true, untrusted content is wrapped in spotlighting XML delimiters
that instruct the LLM to treat the enclosed text as data, not instructions.
quarantine: QuarantineConfigQuarantine summarizer configuration.
embedding_guard: EmbeddingGuardConfigEmbedding anomaly guard configuration.
mcp_to_acp_boundary: boolWhen true, MCP tool results flowing through ACP-serving sessions receive
unconditional quarantine summarization and cross-boundary audit log entries.
This prevents confused-deputy attacks where untrusted MCP output influences
responses served to ACP clients (e.g. IDE integrations).
nli: NliConfigNLI entailment check stage configuration.
secret_masking: SecretMaskingConfigPAAC secret placeholder masking configuration.
Trait Implementations§
Source§impl Clone for ContentIsolationConfig
impl Clone for ContentIsolationConfig
Source§fn clone(&self) -> ContentIsolationConfig
fn clone(&self) -> ContentIsolationConfig
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
source. Read more