agentic-ai-engineering/products/DASHBOARD.md

TAC Products Dashboard

Product	Price	Status	Class	Grade	Agentic layer	Skills	Deploy actions
Product Factory	$99	mvp	3	3	orchestrator-adw	8	patch_skill_from_pr
Plan F3 (Mythos Planning Meta-Skill)	$49	planned	2	3	planner-meta	6	patch_skill_from_pr
Security Gate	$49	planned	2	1	out-loop-policy-gate	6	disable_skill, enable_skill
Verifier Pro	$39	existing-kit	1	4	closed-loop-verifier	5	none
SkillOpt Lab	$39	planned	2	2	eval-driven-improvement	5	run_skill_tests, patch_skill_from_pr
CEO Board	$49	existing-kit	1	7	expert-panel-mental-models	6	none
Observability Cockpit	$29	existing-kit	2	1	out-loop-observability	4	none
Multi-Agent Orchestrator	$49	existing-kit	3	2	orchestrator-workflows	6	none
Task Discipline	$29	existing-kit	1	4	closed-loop-task-prompts	4	none

Generated products

• agent-powered-customer-support-triage-product

Last check

• No checks logged yet

Deploy readiness

• No outcome logged yet

Plan status

• Plans live in plans/<product_id>/plan.md and must validate before build/deploy.
• Amendments are append-only via ## Amendment <UTC timestamp>.

Open risks

• Do not overclaim arbitrary self-improvement; only verified canary/receipts belong here.
• Legacy direct :8098 deploy-webhook is internal/stale for guarded deploy; use :8099/deploy.

Current factory truth

• RSI status: live factory reports 65/65 healthy, cron + skill_health present, factory_watcher active, deploy_route present, and auto_patch_proven=false; rsi_canary recovery evidence is linked at packages/pi-real-engineering/docs/receipts/rsi-proof-20260619T093724Z.md and mirrored at products/receipts/rsi-proof-20260619T093724Z.md.
• Engine factory memory: 25 containers, 6 agents, 63 skills.
• Deploy path must use signed actions through git-proxy :8099/deploy, not raw shell or direct legacy :8098 deploy.

Loop engineering

• Loop 1 agent: ADW planner/builder/tester/reviewer/deployer phases run tools until done.
• Loop 2 verification: plan validation, pytest, verifier events, and receipts gate readiness.
• Loop 3 event-driven: GitHub webhooks, cron skill_health, JSONL events, and dashboard refresh trigger work.
• Loop 4 hill-climbing: SkillOpt/skill_health can request signed patches, but broader auto-patch autonomy stays unproven without degraded-skill recovery receipts.

Next action

python products/product_factory.py "agent-powered customer support triage product"
python products/product_factory.py ladder
uv run --with pytest pytest -q