Secrets logged in CI output
To help you debug a failing deploy, an AI added logging to this GitLab CI job. It "works" — the pipeline now prints useful context. Your security team disables the project's CI minutes and pages you. Why?
Implement
find_secret_leaks(script_lines: list[str], secret_names: list[str]) → list[int]Examples
in
[["echo \"Deploying with config:\"","env","echo \"AWS key is $AWS_SECRET_ACCESS_KEY\"","aws s3 sync ./build s3://acme-prod-site --delete"],["AWS_SECRET_ACCESS_KEY"]]out[2,3]in
[["echo \"Deploying to prod\"","aws s3 sync ./build s3://acme --delete"],["AWS_SECRET_ACCESS_KEY"]]out[]What a strong answer looks like
Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
0:00 of about 13 min
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.
Which questions mattered is sealed until you submit. Telling you now would just be handing over the edge cases.
Run or narrate your approach, then ask the coach.