Safeguard Loop

Roadmap — not yet available. Safeguard Loop is not a new engine or a shipping feature — it's the intended umbrella name and unified view for a continuous cycle that today runs across several separate, already-shipped Safeguard capabilities. This page documents the direction; it is not something you can turn on, and it should not be planned against as a discrete feature yet.

Safeguard already detects issues continuously, prioritizes them, and — for a growing set of cases — fixes and re-verifies them automatically. Right now those steps live in distinct engines: continuous scanning finds things, AutoTriage cuts the noise, Griffin AI recommends and generates fixes, and engines like Auto-Fix and Self-Healing Containers apply and verify them. Safeguard Loop is the vision of naming and presenting that whole detect → prioritize → fix → verify → repeat cycle as one continuous life cycle, rather than as separate tools a team has to mentally stitch together.

Today: the pieces that already exist

Nothing below is new — these are real, shipped capabilities that a future Loop view would tie together into a single life cycle.

Each of these already closes part of the loop on its own. What doesn't exist yet is a single place that shows the cycle running end to end across all of them.

Roadmap: what Loop would add

Safeguard Loop, as envisioned, is a continuous life cycle for security: find issues, fix them, and remediate — including auto-fix — on a repeating basis, rather than as one-off scans and one-off fixes. Directionally, that means:

• A single umbrella view of the cycle — seeing detection, prioritization, and remediation as stages of one continuous loop instead of separate screens per engine.
• Cross-engine progress tracking — following an issue as it moves from detected, to triaged, to a proposed fix, to a verified and promoted fix, regardless of which engine (Package Firewall, Griffin AI, Auto-Fix, Self-Healing Containers, or others) is doing the work at each stage.
• Repeat, not one-shot — reinforcing that the cycle runs continuously rather than stopping once a given fix ships.

To make this concrete: imagine a vulnerable dependency is caught by continuous scanning, deduplicated and prioritized by AutoTriage, given a Griffin AI-generated fix, and turned into a pull request by Auto-Fix. Today, following that one issue end to end means checking three separate places — AutoTriage's findings view, Griffin AI's recommendation, and Auto-Fix's PR status. A Loop view, as envisioned, would show that same issue as a single row moving through detected → triaged → fix proposed → fix verified, without hopping between engines. No such combined view, cross-engine issue ID, or UI exists today — this is an illustration of the goal, not a description of anything built.

None of this changes what the underlying engines above do today. Treat this section as direction, not a specification — the mechanics of a unified Loop view have not been designed or committed to a release yet.

Questions we expect

• Can I turn Loop on today? No. There's no feature flag, setting, or dashboard called Loop anywhere in the product. It's a name and a view being considered for a cycle that already runs across separate, already-shipped engines.
• Does "Loop doesn't exist yet" mean the engines don't work together today? No — Griffin AI already drives both Auto-Fix and Self-Healing Containers directly, and Package Firewall's quarantine step already re-evaluates and releases packages on its own. What's missing is a single view that shows one issue's status across all of those engines at once, not the underlying capabilities.
• Will Loop replace Package Firewall, AutoTriage, Griffin AI, Auto-Fix, or Self-Healing Containers? No, as described — Loop is envisioned as a unifying name and view layered on top of those engines, not a replacement for any of them.
• When will Loop ship, and what will the interface look like? Not committed and not designed yet. This page reflects the product owner's stated direction — a single umbrella view, cross-engine progress tracking, continuous repeat — not a specification or timeline.
• Where can I track an issue end-to-end in the meantime? Through each engine's own view: AutoTriage's correlated finding, Griffin AI's recommendation, Auto-Fix's PR status, the Self-Healing dashboard's heal-cycle status, or Package Firewall's quarantine state. Stitching across them is manual until a unified view exists.

Related

• Griffin AI — the model behind prioritization and remediation generation, available today.
• AutoTriage — cross-scanner deduplication and prioritization, available today.
• AI Remediate — automated fix generation and PR creation, available today.
• Package Firewall — install-time prevention with a quarantine re-evaluation loop, available today.
• Self-Healing Containers — the existing detect-plan-rebuild-promote loop for container images, available today.