[PetrHeinz]

The E2E Linear issue said the same thing up to three times per failed job: the job's errors, then the same errors as ❌ lines under each read-back. Its tables are hard to read in Linear.

• notify_linear_failure.yml adds a failed job's summary or, when the job has none (it stopped before writing one), its errors, without duplicates. Callers that set neither input still get the plain issue.
• The read-back summary lists the counts, and after a timeout what CloudWatch Logs received and forwarded per log group, as bullets instead of tables. The Linear copy drops the heading, since the issue already names the job and links the run.
• A job that stops before the read-back uploads no summary, so its issue shows the step and its errors.

🤖 Generated with Claude Code