Reference

Reason-code reference

Sector8 does not just block actions. It explains why. Every DENY includes a reason code, decision trace ID, evidence hash, and policy version, so security and engineering teams can inspect what happened and reproduce the control. Codes on this page are names the gate actually emits.

What the gate denies

BOLA_VIOLATION

Broken object-level authorisation

What it means: The agent attempted to access or modify an object outside the caller's authorised scope.
Example attempted action: A support agent tries to read or update a customer record belonging to another tenant, account, or ownership boundary.
Why Sector8 denies: The object ID does not match the caller's authorised object scope or tenant boundary.
Evidence captured: Agent IDUser IDObject IDTenant or ownership scopeTool namePolicy versionDecision trace IDEvidence hash
Suggested remediation: Check the object ownership mapping, caller context, and policy scope before retrying.

BFLA_VIOLATION

Broken function-level authorisation

What it means: The agent attempted to use a function or tool outside its permitted role or capability.
Example attempted action: An HR assistant attempts to export payroll data or a support agent attempts to invoke an admin-only CRM update function.
Why Sector8 denies: The function is not permitted for that agent, role, environment, or policy scope.
Evidence captured: Agent IDRoleTool/function nameRequested actionPolicy versionDecision trace IDEvidence hash
Suggested remediation: Update the policy only if the agent truly requires that capability. Otherwise, keep the function denied.

SENSITIVE_PATH_BLOCKED

Sensitive path access blocked

What it means: The agent attempted to read, write, or execute against a protected file path.
Example attempted action: A developer agent attempts to write to /etc/secrets.env, .env, private key files, production configuration, or restricted system paths.
Why Sector8 denies: The path is classified as sensitive and the requested action is not allowed by policy.
Evidence captured: Agent IDUser IDFile pathActionEnvironmentPolicy versionDecision trace IDEvidence hash
Suggested remediation: Use a permitted workspace path or request explicit review for high-impact changes.

TOOL_NOT_REGISTERED

Tool not registered

What it means: The agent attempted to invoke a tool that is not registered for that agent or policy.
Example attempted action: A coding agent calls apply_terraform or an MCP tool that was never registered.
Why Sector8 denies: The tool name is not in the registered set for that caller and policy.
Evidence captured: Tool nameCaller IDPolicy versionDecision trace IDEvidence hash
Suggested remediation: Register the tool and assign it to the correct policy before use.

RATE_LIMIT_EXCEEDED

Rate limit exceeded

What it means: The agent exceeded the allowed rate of proposed actions.
Example attempted action: An agent fires a burst of tool calls above the admission rate limit.
Why Sector8 denies: The request is over the configured rate. Fail-closed.
Evidence captured: Tool nameCaller IDPolicy versionDecision trace IDEvidence hash
Suggested remediation: Slow the agent loop or raise the limit only if the workload is intended.

GUARD_INTERNAL_ERROR

Gate could not evaluate

What it means: Sector8 could not complete evaluation safely. There is no fail-open path.
Example attempted action: An unhandled error on the evaluate path. The runtime must not execute.
Why Sector8 denies: If the gate cannot evaluate, the action does not run.
Evidence captured: Reason codeDecision trace IDEvidence hashPolicy version
Suggested remediation: Retry only after the gate is healthy. Do not execute locally while it is down.

MCP_SHELL_INJECTION_BLOCKED

Shell injection blocked

What it means: The proposed command contained a shell-injection pattern.
Example attempted action: A bash tool call includes command chaining or an injected fragment in the command string.
Why Sector8 denies: The command is treated as unsafe before execution.
Evidence captured: Tool nameDecision trace IDEvidence hashPolicy version
Suggested remediation: Send a single, explicit command. Do not concatenate untrusted input into the shell.

PARAM_INJECTION_DETECTED

Parameter injection detected

What it means: Tool arguments contained an injection pattern.
Example attempted action: A tool argument includes an injected instruction or encoded payload aimed at changing behaviour.
Why Sector8 denies: The arguments fail an admission check. Classification here is a check, not the control.
Evidence captured: Tool nameDecision trace IDEvidence hashPolicy version
Suggested remediation: Strip untrusted input from arguments and retry with the intended parameters only.

PRIVATE_IP_BLOCKED

Private IP blocked

What it means: The agent attempted a network call to a private or link-local address.
Example attempted action: A fetch or HTTP tool targets 127.0.0.1, 10.0.0.0/8, or a link-local address.
Why Sector8 denies: Private destinations are blocked before the call is made.
Evidence captured: Tool nameDecision trace IDEvidence hashPolicy version
Suggested remediation: Call only allowed public destinations, or request a policy that explicitly permits the target.

EXTERNAL_DOMAIN_BLOCKED

External domain blocked

What it means: The agent attempted a network call to a domain that is not allowed.
Example attempted action: A tool posts data to an unapproved external host.
Why Sector8 denies: The destination is outside the allowed domain set.
Evidence captured: Tool nameDecision trace IDEvidence hashPolicy version
Suggested remediation: Use an approved endpoint, or add the domain to policy if it is intended.

Test the gate in shadow mode

Run Sector8 alongside your current agents with enforcement off. See which actions would have been denied, why they were denied, and which policies need to change before enforcement.