Skip to content

fix(classifier): enforce AI enablement interlock#13

Merged
F-e-u-e-r merged 1 commit into
mainfrom
fix/classifier-enable-guard
Jun 22, 2026
Merged

fix(classifier): enforce AI enablement interlock#13
F-e-u-e-r merged 1 commit into
mainfrom
fix/classifier-enable-guard

Conversation

@F-e-u-e-r

Copy link
Copy Markdown
Owner

Problem

ai.enabled was documented as an operational switch but was only displayed. The CLI also did not load config/ai.yaml by default, so an absent explicit --config silently used defaults.

Changes

  • Load config/ai.yaml by default when it exists.
  • When disabled, classifier plan writes a valid empty manifest without README discovery or GitHub credentials.
  • Reject any AI-artifact PR in verify-ai-provenance while classification is disabled.
  • Document the required explicit opt-in and add configuration/CLI regression coverage.

Verification

  • pnpm p3-gate passed: 66 test files, 458 tests, build, and schema drift checks.

Operational use

Before the first executor run, commit a secret-free config/ai.yaml with ai.enabled: true, one executor kind, and a conservative valid budget. The executor token remains external.

Load config/ai.yaml by default, emit no jobs while classification is disabled, and reject attempted AI artifact publication in the provenance gate. Document the explicit opt-in and cover the behavior with CLI and config tests.
@F-e-u-e-r F-e-u-e-r merged commit 38c23e4 into main Jun 22, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant