agentplat▸control runs deployments agents metering + agent
← runs

run_119c3ad8d3027f93

AWAITING_APPROVAL retrigger →
jira-triage v1 · jira-triage-k8s · owner ingest:jira-new-ticket (api) · via jira-new-ticket · cloud-engineering

input

{
  "request": "Triage Jira ticket MZCLD-3754: Code Reviewer discovery (CodeRabbit, @claude, or similar)",
  "references": [
    "https://mozilla-hub.atlassian.net/browse/MZCLD-3754"
  ],
  "sourceContext": {
    "issue": "MZCLD-3754"
  }
}

summary

MZCLD-3754 is a Backlog epic (0 comments, unassigned) asking for a discovery to pick a code-review tool (CodeRabbit, @claude GitHub App, or similar) to pilot across Mozilla repos generally, distinct from the narrower IaC-specific Claude review effort already underway in MZCLD-3215. The ticket is well-scoped but has two concrete pre-work blockers (owner assignment and GitHub App/org policy work by jbuck) that should be resolved before discovery can start. I could not fetch MZCLD-3215 for cross-reference detail due to a tool access-policy restriction.

findings

  • Ticket is unowned and unstarted
  • Clear, output-driven scope: this is a discovery epic, not a commitment to any specific tool
  • Hard blocker identified: GitHub App installation and org-level policy must be done by Jon Buckley (jbuck) before any pilot can run
  • Explicit out-of-scope boundaries defined, including overlap with MZCLD-3215
  • Related/parent context ticket MZCLD-3215 (the narrow IaC Claude-review case that motivated this epic) could not be retrieved due to an access-policy restriction on the read tool, so its current status/PR examples couldn't be verified in this triage
  • No Confluence runbook or design doc is referenced in the ticket text, so no additional design context was available to pull

proposed actions

jira.comment.postrisk low - informational triage comment proposing next steps and clarifying blockers; no ticket fields changed, no commitments made on behalf of any team.
MZCLD-3754
Triage notes: This epic is well-scoped (discovery only, adopt/don't-adopt output, explicitly not committing to a vendor) but currently has no owner and no visible next steps. Before discovery work can start, two things need to happen: (1) confirm an owner/DRI — ticket notes possible involvement from Krupa's team, TBC — and (2) loop in Jon Buckley (jbuck) early, since GitHub App installation and org-level policy on the GitHub Enterprise org is a hard prerequisite for any pilot ("Nothing gets piloted without that"). Recommend also cross-checking with MZCLD-3215 once discovery scope is defined, to make sure the set of pilot repos/tools doesn't duplicate the IaC-specific Claude review work already in progress there. Suggest moving this out of pure Backlog once an owner is assigned and a candidate repo list + tool shortlist (CodeRabbit, @claude GitHub App, others) is drafted.

tool calls

[
  {
    "capability": "jira.issue.read",
    "args": {
      "key": "MZCLD-3754"
    },
    "result": {
      "key": "MZCLD-3754",
      "summary": "Code Reviewer discovery (CodeRabbit, @claude, or similar)",
      "status": "Backlog",
      "description_text": "Background Human review of infrastructure changes does not scale reliably. Atlantis plans and large Helm diffs regularly get approved without anyone genuinely understanding them, and we have real incidents to show for it \u2014 see MZCLD-3215, which walks through the argument and links several PRs that were reviewed and still caused outages. MZCLD-3215 is already proving the narrow case: automated Claude review on IaC repos. This epic is the broader question \u2014 is there an off-the-shelf code reviewer we should be running across Mozilla repos generally, and if so, which one? Goal Run a discovery to determine which code reviewer to use \u2014 CodeRabbit, the  @claude  GitHub App, or something similar \u2014 try it on an agreed set of repos, and come back with an adopt / don't-adopt recommendation. Which tool, which repos, and how far to take it are  outputs  of this discovery, not inputs. The point of the epic is to answer those questions rather than assume them. Ownership and who's involved Deliberately unassigned for now. This  may involve Krupa's team  \u2014 to be confirmed. Jon Buckley (jbuck) is needed for the GitHub side : installing the GitHub App and the org-level policy on the GitHub Enterprise org. Nothing gets piloted without that. Out of scope Productionising or committing Mozilla to a vendor off the back of the discovery, and anything already covered by MZCLD-3215.",
      "comment_count": 0,
      "comments": []
    }
  },
  {
    "capability": "jira.issue.read",
    "args": {
      "key": "MZCLD-3215"
    },
    "result": {
      "error": "read blocked by policy: resource not in run subject",
      "args": {
        "key": "MZCLD-3215"
      }
    },
    "blocked": true
  }
]

usage

{
  "model": "claude-sonnet-5",
  "input_tokens": 5738,
  "output_tokens": 2134,
  "estimated_cost_usd": 0.049224000000000004
}

approvals

[]

log

launched k8s job run-run-119c3ad8d3027f93 in ns/mzcld-demo-dev image=us-docker.pkg.dev/moz-fx-mzcld-demo-prod/mzcld-demo-prod/agentplat:latest
runner started
live reasoning model=claude-sonnet-5
POLICY: blocked read jira.issue.read args={'key': 'MZCLD-3215'} not in run subject
awaiting approval for 1 action(s)