Ecos Permission Management

Emasoft Chief of Staff - Permission Management Skill

Overview

Permission management is a critical governance function of the Chief of Staff. Before performing certain operations that affect agent resources or system state, ECOS must request approval from the Assistant Manager (EAMA), who serves as the user's representative. This skill teaches you how to request approvals, track pending approvals, handle timeouts, and maintain audit trails.

Prerequisites

Before using this skill, ensure:

1. Permission registry is accessible
2. Agent roles are defined
3. EAMA approval workflow is available for escalations

Instructions

1. Identify permission change needed
2. Verify requester authorization
3. Apply permission change
4. Log the change and notify affected agents

Output

What Is Permission Management?

Permission management is the process of obtaining authorization before executing privileged operations. The Chief of Staff must not unilaterally spawn agents, terminate agents, hibernate agents, wake agents, or install plugins without proper approval from the manager (EAMA) unless operating under an explicit autonomous directive.

Key principle: ECOS proposes, EAMA approves. The user (via EAMA) maintains control over resource-consuming operations.

When Approval Is Required

+-------------------------------------------------------------+
|                 APPROVAL REQUIRED OPERATIONS                 |
+-------------------------------------------------------------+
|  AGENT SPAWN       - Creating new agent instances            |
|  AGENT TERMINATE   - Permanently stopping agent execution    |
|  AGENT HIBERNATE   - Suspending agent to conserve resources  |
|  AGENT WAKE        - Resuming hibernated agent               |
|  PLUGIN INSTALL    - Installing new Claude Code plugins      |
+-------------------------------------------------------------+

Exception: If the manager has issued an autonomous operation directive, ECOS may proceed without approval but must notify EAMA after the operation completes.

The Approval Workflow

ECOS                           EAMA                          USER
  |                              |                              |
  |  1. Request approval         |                              |
  |----------------------------->|                              |
  |                              |  2. Present to user          |
  |                              |----------------------------->|
  |                              |                              |
  |                              |  3. User decides             |
  |                              |<-----------------------------|
  |  4. Receive response         |                              |
  |<-----------------------------|                              |
  |                              |                              |
  |  5. Execute or abort         |                              |
  |  6. Log to audit trail       |                              |

For ACK timeout policy and message retry procedures, see the ecos-notification-protocols skill.

Core Procedures

PROCEDURE 1: Request Approval from Manager

When to use: Before executing any agent lifecycle operation or plugin installation.

Steps: Identify operation type, compose approval request, send via AI Maestro, await response, handle decision.

See references/approval-request-procedure.md for complete documentation:

• 1.1 What is an approval request
• 1.2 When to request approval (spawn, terminate, hibernate, wake, plugin)
• 1.3 Approval request procedure (identification, justification, composition, transmission, awaiting)
• 1.4 Request message format
• 1.5 Examples
• 1.6 Troubleshooting

PROCEDURE 2: Track Pending Approvals

When to use: When managing multiple operations requiring approval, when checking status of pending requests.

Steps: Register new requests, monitor response status, handle multiple concurrent requests, update tracking on resolution.

See references/approval-tracking.md for complete documentation:

• 2.1 What is approval tracking
• 2.2 Tracking data structure
• 2.3 Tracking procedure (registration, monitoring, concurrent handling, resolution)
• 2.4 State file format
• 2.5 Examples
• 2.6 Troubleshooting

PROCEDURE 3: Handle Approval Timeouts and Escalation

When to use: When no response received within timeout period, when urgent operation is blocked.

Steps: Send reminder notifications, send urgent notification, determine proceed or abort, log escalation event.

See references/approval-escalation.md for complete documentation:

• 3.1 What is approval escalation
• 3.2 Escalation triggers (60s reminder, 90s urgent, 120s proceed/abort)
• 3.3 Escalation procedure
• 3.4 Autonomous operation mode
• 3.5 Examples
• 3.6 Troubleshooting

Approval Types

See references/approval-types-detailed.md for justification requirements and detailed decision options.

Audit Trail Requirements

All approval operations must be logged:

audit_trail:
  - timestamp: "ISO-8601"
    operation: "spawn|terminate|hibernate|wake|plugin_install"
    target: "agent_name_or_plugin_name"
    request_id: "uuid"
    requested_at: "ISO-8601"
    decision: "approved|rejected|modified|timeout_proceed|timeout_abort"
    decided_at: "ISO-8601"
    decided_by: "eama|autonomous|timeout"
    justification: "reason provided"
    modifications: null | {changes}
    escalation_count: 0|1|2|3

Audit file location: docs_dev/audit/ecos-approvals-{date}.yaml

Task Checklist

Copy this checklist and track your progress:

• Understand when approval is required
• Learn PROCEDURE 1: Request approval from manager
• Learn PROCEDURE 2: Track pending approvals
• Learn PROCEDURE 3: Handle approval timeouts
• Practice sending an agent spawn approval request
• Practice tracking multiple pending approvals
• Practice handling a timeout scenario
• Verify audit trail is properly maintained

Examples

For complete examples with expected responses, see references/examples.md:

• Example 1: Requesting approval to spawn an agent
• Example 2: Requesting approval to terminate an agent
• Example 3: Handling approval timeout (reminder, urgent, proceed)
• Example 4: Operating in autonomous mode (post-operation notification)

Operational Procedures

Step-by-step runbooks for executing each permission management operation. Use these when performing the actual procedures described above.

Request Approval (references/op-request-approval.md)

Detailed step-by-step runbook for requesting approval from EAMA before executing privileged operations (spawn, terminate, hibernate, wake, plugin install).

• When to Use: Before any agent lifecycle operation or plugin installation
• Step 1: Identify Operation Type (spawn, terminate, hibernate, wake, plugin_install)
• Step 2: Generate Request ID (UUID)
• Step 3: Compose Approval Request (with justification)
• Step 4: Send Request via AI Maestro
• Step 5: Register Pending Approval
• Step 6: Await Response (poll every 10s, max 120s)
• Step 7: Handle Decision (approved, rejected, modified)
• Request Message Format, Examples, Error Handling

Track Pending Approvals (references/op-track-pending-approvals.md)

Detailed step-by-step runbook for maintaining tracking of all outstanding approval requests to manage multiple concurrent operations.

• When to Use: Managing multiple approval requests, checking status, generating reports, handling escalation timing
• Step 1: Initialize Tracking File (docs_dev/pending-approvals.json)
• Step 2: Register New Request
• Step 3: Check Pending Requests Status
• Step 4: Poll for Responses
• Step 5: Check for Timeouts (60s reminder, 90s urgent)
• Step 6: Update Tracking on Resolution
• Step 7: Generate Status Report
• Tracking State Schema, Examples, Error Handling

Handle Approval Timeout (references/op-handle-approval-timeout.md)

Detailed step-by-step runbook for handling situations where approval requests do not receive timely responses, including reminders, escalation, and proceed/abort decisions.

• When to Use: Approval pending >60s, urgent escalation needed at >90s, maximum timeout (120s) reached
• Step 1: Check Request Age
• Step 2: Send Reminder at 60 Seconds
• Step 3: Send Urgent Notification at 90 Seconds
• Step 4: Handle Timeout at 120 Seconds (proceed or abort based on operation type)
• Default Timeout Actions (spawn/wake proceed; terminate/hibernate/plugin_install abort)
• Autonomous Mode, Escalation Timeline, Error Handling

Error Handling

Key Takeaways

1. Always request approval - Never execute privileged operations without approval
2. Use proper message format - Include all required fields in approval requests
3. Respect the timeout - 2 minutes max wait, with notifications at 60s and 90s
4. Log everything - Maintain complete audit trail for all approval operations
5. Handle autonomous mode - When directive given, skip approval but still notify
6. Track pending requests - Maintain state for all outstanding approvals

Plugin Prefix Reference

Resources

• Approval Request Procedure
• Approval Tracking
• Approval Escalation
• Approval Types Detailed
• Examples
• RULE 14 Enforcement - User requirements are immutable
• Approval Workflow Engine - Complete approval workflow engine procedures

Version: 1.0 Last Updated: 2025-02-03 Target Audience: Chief of Staff Agents Difficulty Level: Intermediate

Converted and distributed by TomeVault — claim your Tome and manage your conversions.