Skip to content

edge-scrum: chai-bot report skills for the weekly health report (OCPEDGE-3121) - #331

Open
Neilhamza wants to merge 6 commits into
openshift-eng:mainfrom
Neilhamza:OCPEDGE-3121-chai-report-skills
Open

Neilhamza wants to merge 6 commits into
openshift-eng:mainfrom
Neilhamza:OCPEDGE-3121-chai-report-skills

Conversation

@Neilhamza

@Neilhamza Neilhamza commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

What's added

  • chai/sprint-health, chai/release-health, chai/weekly-health-report: chai-bot coordinator skills for the weekly report of edge-context decision 0018 (no sub-agents, no prompts, team level only, missing data listed and never estimated).
  • bin/chai_jira_adapter.py: query_jira text → transform input, checked against real chai-bot output (fixtures are anonymized copies); key-ordered paging check (per project, since Jira's key < is project-scoped); display-name roster match; get_jira_issue link parser; feature spikes via linkedIssues("<F>", "is blocked by"); team-roster.md parser (Rover email, "Not counted" excluded).
  • bin/chai_report.py: mode, capacity (velocity × availability, else "not PTO-adjusted"; in release health the current sprint counts only its remaining working days), phase → section selection, Slack text. Unpointed work is a count, never an SP estimate (run-checks.py --no-hidden-scope-estimate); team-level actions go to the Scrum Master when edge-context names one.
  • scrum_context.py / chai/scrum-context: per-sprint availability up to branch cut; "Not counted" members excluded. edge-scrum 1.6.0.

Tests

516 pass (115 new) with python3 -m pytest plugins/edge-scrum/bin/tests -q; markdownlint clean on the changed files.

Dry run (2026-11-08, live Jira via mcp-atlassian, local only)

  • Same data on both paths: current sprint 141 issues / 190 SP, burndown, unpointed and epic-less issues; 5.1 has 17 features, scope 82 SP. With full sprints and the same roster, chai matched Claude Code's capacity and cut line exactly.
  • Capacity differs by design: Claude Code release-planning counts the current sprint in full (2 sprints); chai counts only its remaining working days (0 of 15 on the run day), so 80 SP vs 82 SP scope, one feature below the cut line.
  • Velocity 80.3 vs 83.7 (capacity 161 vs 167 SP, one more feature at risk): one engineer is in the local .roster.json but not in edge-context team-roster.md. With the same roster the chai path matches exactly.
  • Blocked or stalled 7 vs 23, missing test plan spikes 4 vs 8: pre-existing Claude Code bug. mcp-atlassian nests a linked issue's status and type under fields, so extract_blocked_by counts closed blockers and transform-features.py finds no linked spikes; the midpoint analyzer also lists issues that are already done. The chai figures are correct; fix as a follow-up.
  • Not compared: availability (calendar not readable here, both "not PTO-adjusted"), Product Pages (anchors taken from the board), acceptance criteria (left out in chai).

Follow-up

  • Follow-up: OCPEDGE-3239 (JSON swap, blocked by CHAI-1248)
  • Validate in a chai-bot DM after the skills are registered (plugins/edge-scrum/chai/).
  • Error and cut-off query_jira pages have not been seen yet; their handling is still an assumption.

Summary by CodeRabbit

  • New Features
    • Added team-level sprint, release, and refinement health reports covering capacity, availability, risks, blockers, and action items.
    • Weekly reports can include a concise summary and ordered replies, with action items assigned to mapped owners.
    • Reports show availability for the current sprint and later applicable sprints, and can identify unpointed scope without estimating it.
    • Added workflows for gathering Jira and roster information and checking data completeness before reporting.
  • Documentation
    • Documented reporting workflows, availability rules, and how incomplete data is handled.
  • Chores
    • Updated the edge-scrum plugin version to 1.6.0.

Loading
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. ready-for-human-review Indicates a PR has been reviewed by automated tools and is ready for human review

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants