Governance

Policies

Review fictional policy declarations, scope, and agent assignment.

Frontend prototype. Policy summaries are fictional declarations only. No rule is evaluated, enforced, authorized, or persisted.

4 of 4 fictional policies

Policy inventory

Active	Connector least privilege Declares that agent tools may request only named connector operations and scopes. P-307 · v5 View readable rule summary Deny operations not explicitly declared. Never expose provider clients or credentials to agents. Material scope changes require review. Display summary only. No policy is evaluated or enforced.	Access All connector operations	Connector Health Sentinel Calendar Briefing Agent
Active	Duplicate run safeguard Describes a future safeguard against overlapping scheduled runs. P-402 · v2 View readable rule summary Check the active-run key before queueing. Do not imply that this UI performs the check. Record a normalized blocked outcome in a future runtime. Display summary only. No policy is evaluated or enforced.	Execution Scheduled agent runs	Workspace-wide declaration
Paused	Evidence retention declaration Declares fictional retention metadata for governance evidence. P-118 · v3 View readable rule summary No enforcement exists in this frontend prototype. Retention dates must be explicit metadata. Expired content must not be presented as available. Display summary only. No policy is evaluated or enforced.	Retention Governance evidence metadata	Policy Digest Agent
Active	External communications approval Requires human confirmation before a fictional high-impact external communication could proceed. P-214 · v8 View readable rule summary Critical requests require step-up confirmation. Model output never authorizes the action. Approval evidence must reference the originating run. Display summary only. No policy is evaluated or enforced.	Approval Outbound communication drafts	Support Draft Agent

Connector least privilege
P-307 · v5
Active
Declares that agent tools may request only named connector operations and scopes.
Type
Access
Scope
All connector operations
Assigned agents
Connector Health Sentinel
Calendar Briefing Agent
View readable rule summary
Deny operations not explicitly declared.
Never expose provider clients or credentials to agents.
Material scope changes require review.
Display summary only. No policy is evaluated or enforced.
Duplicate run safeguard
P-402 · v2
Active
Describes a future safeguard against overlapping scheduled runs.
Type
Execution
Scope
Scheduled agent runs
Assigned agents
Workspace-wide declaration
View readable rule summary
Check the active-run key before queueing.
Do not imply that this UI performs the check.
Record a normalized blocked outcome in a future runtime.
Display summary only. No policy is evaluated or enforced.
Evidence retention declaration
P-118 · v3
Paused
Declares fictional retention metadata for governance evidence.
Type
Retention
Scope
Governance evidence metadata
Assigned agents
Policy Digest Agent
View readable rule summary
No enforcement exists in this frontend prototype.
Retention dates must be explicit metadata.
Expired content must not be presented as available.
Display summary only. No policy is evaluated or enforced.
External communications approval
P-214 · v8
Active
Requires human confirmation before a fictional high-impact external communication could proceed.
Type
Approval
Scope
Outbound communication drafts
Assigned agents
Support Draft Agent
View readable rule summary
Critical requests require step-up confirmation.
Model output never authorizes the action.
Approval evidence must reference the originating run.
Display summary only. No policy is evaluated or enforced.

Session-only action

Active

Connector least privilege

Declares that agent tools may request only named connector operations and scopes.

P-307 · v5

View readable rule summary

Deny operations not explicitly declared.
Never expose provider clients or credentials to agents.
Material scope changes require review.

Display summary only. No policy is evaluated or enforced.

Access

All connector operations

Active

Duplicate run safeguard

Describes a future safeguard against overlapping scheduled runs.

P-402 · v2

View readable rule summary

Check the active-run key before queueing.
Do not imply that this UI performs the check.
Record a normalized blocked outcome in a future runtime.

Display summary only. No policy is evaluated or enforced.

Execution

Scheduled agent runs

Workspace-wide declaration

Paused

Evidence retention declaration

Declares fictional retention metadata for governance evidence.

P-118 · v3

View readable rule summary

No enforcement exists in this frontend prototype.
Retention dates must be explicit metadata.
Expired content must not be presented as available.

Display summary only. No policy is evaluated or enforced.

Retention

Governance evidence metadata

Policy Digest Agent

Active

External communications approval

Requires human confirmation before a fictional high-impact external communication could proceed.

P-214 · v8

View readable rule summary

Critical requests require step-up confirmation.
Model output never authorizes the action.
Approval evidence must reference the originating run.

Display summary only. No policy is evaluated or enforced.

Approval

Outbound communication drafts

Support Draft Agent