Platform Readiness Report — Mission Zero Enablement

Date: 2026-03-28 Status: Phases 1-3 Complete, Deployed, Migration Applied Commits: eff58ec, 758faa7, d3ab771

---

What Was Done

Mission Zero identified critical platform gaps preventing agents from self-managing. This report summarises the 3-phase build that addresses those gaps.

Phase 1: Agent Analytics Tools (Complete)

Problem: 40+ analytics API endpoints existed but only 5 were exposed as platform tools. Agents like ATLAS and SENTINEL had no way to self-monitor.

Solution: 11 new platform tools wrapping real database queries:

Tool	What It Returns
platform_get_success_rate	Overall agent success rate + 7-day trend
platform_get_completion_time	Avg task completion time + 24h comparison
platform_get_error_rates	Error breakdown by agent type with severity
platform_get_queue_depth	Pending/running task counts
platform_get_efficiency_score	Composite 0-100 score with A-D grade
platform_get_cost_per_execution	Cost trends with daily breakdown
platform_get_peak_hours	24h usage pattern with peak/medium/low
platform_get_bottlenecks	Failure rates, queue buildup, slow executions
platform_get_predictive_alerts	Cost spikes, capacity warnings, rate limits
platform_get_agent_ranking	Agents ranked by composite performance score
platform_get_sla_compliance	Task completion + response time vs SLA targets

Impact: Agents can now answer "how are we performing?", "what's costing the most?", "any problems?" without human intervention.

---

Phase 2: KPI Dashboard Widgets (Complete)

Problem: The command centre had 8 operational widgets but zero KPI visibility for cost tracking, agent performance, playbook health, or approval status.

Solution:

4 new widgets:

• Cost Tracker — Total spend, daily sparkline chart, top 3 agent spenders, period-over-period change %

• Agent Performance — Color-coded horizontal bars (green >90%, yellow 70-90%, red <70%) ranked by success rate

• Playbook Metrics — Table with runs, success %, avg duration per playbook

• Approval Gates — Pending count, avg approval time, waiting mission list

Widget Picker (new capability):

• Dashboard refactored from static list to a widget registry pattern

• Click Customize -> "Widgets" popover shows all 12 widgets with show/hide toggles

• Drag to reorder still works

• Hidden state persists in localStorage

• New widgets auto-appear on existing installs (merge logic)

• Ready for a future widget marketplace

Backend: New /api/kpi router with 4 endpoints, all workspace-scoped.

---

Phase 3: Governance & Blueprints (Complete)

Problem: workspace.plan_limits JSONB exists but was NEVER READ. Agent.is_approved exists but was NEVER ENFORCED. Zero budget tracking on missions. No quality standards.

Solution:

New table: agent_blueprints

• Per-workspace governance rules

• Configurable: min_tools, require_system_prompt, max_budget_per_run, required_tags, allowed_models

• One blueprint can be set as workspace default

New columns on orchestration_runs:

• budget_config — {max_cost, max_tokens, alert_at_pct}

• budget_spent — {cost, tokens, api_calls}

6 new platform tools:

Tool	Permission	Purpose
platform_list_blueprints	read	List workspace blueprints
platform_get_blueprint	read	Get blueprint details
platform_create_blueprint	write	Create governance rules
platform_update_blueprint	write	Update rules
platform_validate_agent	read	Check agent against blueprint (pass/fail + reasons)
platform_check_budget	read	Mission budget status (ok/warning/exceeded)

Blueprint Validator Service:

• Validates agents against configurable rules

• Returns specific failures ("Agent has 0 tools, minimum is 2") and warnings ("Missing recommended tags")

• Budget checker returns remaining cost/tokens with alert thresholds

Frontend: AgentReadinessBadge

• Green shield = all checks passed

• Yellow shield = warnings

• Red shield = failures

• Tooltip shows specific issues

---

Platform Tool Inventory

Total platform tools: 97 across 19 domains:

Domain	Tools	Examples
Agents	6	list, get, create, update, delete
Playbooks/Recipes	10	CRUD + steps + execute
Analytics (base)	5	LLM usage, costs, workspace stats
Analytics (enhanced)	11	Success rate, bottlenecks, SLA, rankings
Governance	6	Blueprints, validation, budgets
Board Tasks	6	CRUD + assign + status
Marketplace	9	Browse agents/plugins/skills, install
Monitoring	5	Logs, Loki, Prometheus, alerts, health
Documents	3	List, delete, reprocess
Workspace	7	Info, memory, connected apps, store
Tools/LLMs	3	List tools, LLMs, datasources
Search	2	Chat history, memory search
Scheduling	3	Schedule, list, cancel
Reports	2	Submit, get latest
Assignments	3	Tool, skill, plugin assignment + heartbeat
Field	3	Query, inject, stability
Blog	4	Publish, list, get, update
Missions	3	Create, list, get
Workspace I/O	6	read_file, write_file, list_dir, grep, exec, git

---

Mission Zero Gaps — Status

Gap Identified	Status	Notes
Analytics not exposed to agents	FIXED	11 new tools (Phase 1)
No KPI visibility for humans	FIXED	4 widgets + picker (Phase 2)
No governance/quality standards	FIXED	Blueprints + validator (Phase 3)
Budget fields exist but never read	FIXED	budget_config/budget_spent + check tool
is_approved never enforced	PARTIAL	Blueprint validation warns, doesn't block yet
Tool discovery for 880+ Composio actions	ALREADY DONE	platform_list_tools with search exists
Channel abstraction	ALREADY DONE	Templates use types, not specific channels
Agent config gaps (0 tools, no prompts)	NOW DETECTABLE	platform_validate_agent flags these

---

What Auto Can Now Do

1. Self-monitor: "What's our success rate?" -> real data from platform_get_success_rate

2. Investigate problems: "Any bottlenecks?" -> failure rates, queue buildup, slow execution detection

3. Track costs: "Are costs going up?" -> daily breakdown with trend analysis

4. Rank agents: "Who's the top performer?" -> composite scoring across success, speed, volume

5. Check SLAs: "Are we meeting targets?" -> completion rate + response time vs thresholds

6. Set standards: "Create a blueprint requiring prompts and 2 tools" -> governance enforcement

7. Validate agents: "Is SCOUT ready?" -> specific pass/fail with actionable reasons

8. Monitor budgets: "How much budget is left on this mission?" -> remaining cost/tokens with alerts

9. Predict issues: "Any upcoming capacity problems?" -> cost spikes, agent capacity, rate limits

---

Verification Checklist

• Ask ATLAS: "What's our agent success rate?" — should return real % from executions

• Ask SENTINEL: "Any bottlenecks?" — should detect queue/failure/speed issues

• Open Activity -> Command Centre -> Customize -> Widgets — should show 12 widgets with toggles

• Cost Tracker widget shows real spend data

• Agent Performance widget shows color-coded bars

• Ask Auto: "Create a blueprint requiring system prompts and at least 2 tools"

• Ask Auto: "Validate agent SCOUT against the blueprint" — should flag missing tools/prompt

• Ask Auto: "Do you have everything you need to self-manage?" — the real test

---

Next Steps

1. Google Analytics integration — GA4 added to workspace, needs frontend snippet + platform tool + KPI widget

2. Budget enforcement — Currently advisory (check_budget returns status). Next: auto-pause missions when exceeded

3. Blueprint enforcement in coordinator — Currently tools-only. Next: coordinator checks before dispatch, logs warnings

4. Widget marketplace — Registry pattern is ready for user-installed widgets