[lucidprogrammer]

Hosted run: CI 40181786646 — green, 33 min 14 s.

Commit (on dev cd75f16)

• 0f9be063 ci: install ripgrep for scheduled flake detection and release workflows

What changed

crates/contextdb-cli/tests/read_classification_staged.rs refuses to run without jq and rg, so the command-classification proof can never pass silently. ci.yml installs ripgrep; four workflows that run the full workspace suite did not, so that test failed on every attempt there (#7, and the workspace job of every nightly). This adds the same install step to flake-hunt.yml, release.yml, release-dry-run.yml and the nightly integration.yml slice job, and pins the new integration.yml command in scripts/ci-integration.py's workflow contract. jq is already on ubuntu-latest.

Proof

• The PR run does not include that test (PR bundles run with CI_INTEGRATION_DEFAULT=false). With ripgrep present it passes: cargo test --locked -p contextdb-cli --test read_classification_staged → 1 passed.
• Offline CI controls ci-integration-tests.py (16), ci-inventory-tests.py (14), ci-required-tests.py (11) pass.
• Not fixed here: the nightly production-layout jobs fail to compile acceptance, acceptance_sync and sql_surface without test seams (known, .github/CI.md).

Cross-family review (Grok)

First pass STEER: the nightly integration.yml slice also lacked ripgrep — added, with its contract line. Second pass: no findings.

Fixes #7.