v0.9.2

✨ Added

• The JSON envelope now carries Jira's own trace id: the Atl-Traceid / X-ARequestId response header is captured as meta.upstream_request_id and on each error entry, so a Jira-side failure can be quoted directly to Atlassian support. The existing meta.request_id is generated locally and has no server-side meaning.

🪲 Fixed

• Human output now strips ANSI escape and control bytes from Jira-controlled text (summaries, descriptions, error messages), so a crafted issue or error can no longer inject color codes or corrupt your terminal.
• jira agent schema now binds the comment input schema to the comment add/comment edit leaves (they previously reported no input schema) and publishes output schemas for the read/list commands — comment, attachment, link, worklog, transition, boards, and cache — so agents can discover response shapes without trial calls.
• issue attachment download --to is now confined to the working directory: a .. traversal or an absolute path outside it is rejected before any HTTP call, so an agent (or a crafted instruction) can no longer write downloads outside the tree it was launched in. A new security_posture section in jira agent guide documents the threat model.

📜 Commits

• f58f060 fix(cli): sanitize Jira-controlled text in human terminal output
• aca151c fix(schema): bind comment input schema to leaves and fill sparse output schemas
• 1bb41ce fix(cli): confine attachment download paths to the working directory
• 01adc7b feat(cli): capture Jira upstream trace id in envelope meta and errors

Full Changelog: v0.9.1...v0.9.2