[bmad-onboard S1] Onboard sre-lead persona #1307

Activity

1. don-petry

   added

   dev-leadFor dev-lead agent pickupFor dev-lead agent pickup

   on Jul 18, 2026
2. github-actions

   added

   enhancementFeature requestsFeature requests

   on Jul 18, 2026
3. github-actions commented on Jul 18, 2026

   

   github-actionsBot

   on Jul 18, 2026 – with GitHub Actions

   Contributor

   More actions

   Thanks for filing this and for the clear task breakdown! Since this is implementation work rather than a question, one thing worth confirming: once #1306 merges and frameworks/bmad-bgreat-suite/ is available, is there any preference on which existing persona directory (qa-lead vs scrum-master) should serve as the primary template to diff against for the persona.yml structure?
4. don-petry

   added

   dev-leadFor dev-lead agent pickupFor dev-lead agent pickup

   and removed

   dev-leadFor dev-lead agent pickupFor dev-lead agent pickup

   on Jul 18, 2026
5. don-petry commented on Jul 18, 2026

   

   don-petry

   on Jul 18, 2026

   CollaboratorAuthor

   More actions

   Dev-Lead Implementation Plan

   Issue: #1307 — [bmad-onboard S1] Onboard sre-lead persona

   Scope

   Onboard sre-lead as a draft advisory persona wrapping the vendored bgr-agent-morgan-sre agent, following the exact pattern of qa-lead/scrum-master (manifest + advisory prompt + README + seed evals).

   Implementation Checklist

   • evals/sre-lead/dev/cases.jsonl — ≥2 synthetic/de-identified SRE cases
   • evals/sre-lead/holdout/cases.jsonl — ≥2 synthetic/de-identified SRE cases (disjoint ids)
   • personas/sre-lead/persona.yml — status: draft; framework-agent layer → frameworks/bmad-bgreat-suite/src/agents/bgr-agent-morgan-sre; framework.name: bmad-bgreat-suite; vendor_pin = SHA ae8914e84b87 from VENDOR.md; skill: bgr-agent-morgan-sre; address.handle: petry-projects/sre-lead; advisory default_mode; opt_out_label: sre-lead:hands-off; evals.path: evals/sre-lead/; canary.agent: sre-lead
   • prompts/sre-lead/advisory.md — copy of prompts/qa-lead/advisory.md, specialized to SRE domain (observability, incident response, SLO/SLI, reliability, chaos, postmortems); marker <!-- persona:sre-lead -->; marker rule + no-literal-handle rule verbatim
   • personas/sre-lead/README.md — what it is, org team handle addressing, status draft, overlap: none significant

   Tests to Write

   • Seed eval cases act as the test fixtures; verified via validate-cases.py (directory mode)
   • Manifest verified via validate-personas.py (5→6 personas)

   Test Command

   python3 personas/validate-personas.py personas and python3 evals/validate-cases.py evals and bash .dev-lead/scripts/dev-lead-lint.sh and markdownlint-cli2 personas/sre-lead/README.md
6. don-petry commented on Jul 18, 2026

   

   don-petry

   on Jul 18, 2026

   CollaboratorAuthor

   More actions

   Dev-Lead: Implementation Complete

   Plan Execution

   • evals/sre-lead/dev/cases.jsonl — 3 synthetic SRE cases (no golden signals, raw-threshold alert, well-instrumented negative)
   • evals/sre-lead/holdout/cases.jsonl — 4 synthetic SRE cases (critical-path no-SLO, region-failover no-DR, blameful postmortem, prod chaos no-safety); ids disjoint from dev
   • personas/sre-lead/persona.yml — status: draft; framework-agent → frameworks/bmad-bgreat-suite/src/agents/bgr-agent-morgan-sre; framework.name: bmad-bgreat-suite; vendor_pin: ae8914e84b87 (matches VENDOR.md); skill: bgr-agent-morgan-sre; address.handle: petry-projects/sre-lead; advisory default_mode; opt_out_label: sre-lead:hands-off; evals.path: evals/sre-lead/; canary.agent: sre-lead
   • prompts/sre-lead/advisory.md — copied from prompts/qa-lead/advisory.md, specialized to SRE (observability, incident response, SLO/SLI, DR, chaos, postmortems); first body line <!-- persona:sre-lead -->; marker rule + no-literal-handle rule kept verbatim
   • personas/sre-lead/README.md — what it is, org-team-handle addressing (never bare @sre-lead), status draft, overlap: none significant

   Test Results

   $ python3 personas/validate-personas.py personas
   OK: 6 persona manifest(s) valid, invariants hold.          # 5 -> 6
   
   $ python3 evals/validate-cases.py evals
   OK: 10 skill(s), 82 case(s) across dev+holdout, no cross-split id overlap.
   
   $ bash .dev-lead/scripts/dev-lead-lint.sh
     [lint] shellcheck on 95 file(s)...
     [lint] validating 4 agent profile(s)...
     [lint] validating .github/CODEOWNERS...
     [lint] all checks passed
   
   $ npx markdownlint-cli2 personas/sre-lead/README.md
   Summary: 0 issues in 0 files
   

   Files Changed

   • personas/sre-lead/persona.yml: new draft advisory manifest wrapping the vendored bgr-agent-morgan-sre agent
   • prompts/sre-lead/advisory.md: read-only/sentinel/marker advisory contract, SRE-specialized
   • personas/sre-lead/README.md: persona doc (role naming, addressing, draft promotion checklist, overlap note)
   • evals/sre-lead/dev/cases.jsonl, evals/sre-lead/holdout/cases.jsonl: seed held-out eval sets

   Notes

   • Eval expected shape (risk_tier/escalate/recommend) mirrors the existing qa-lead cases; CI runs validate-cases.py in directory mode over the committed evals/ tree (id/split hygiene) — the file-mode case.schema.json is triage-specific and not applied here, exactly as for qa-lead.
   • Role-not-person (§1.6) upheld: id/name/title are sre-lead / SRE Lead / Site Reliability & Operations Advisor; the upstream agent is referenced only by its technical framework.skill (bgr-agent-morgan-sre).
   • Per issue scope: did not create the org team and did not register in canary-rings.json (draft).
   • Phase 2 (TDD): the seed eval cases serve as the persona's test fixtures; validation is via the two validators. No executable behavior/scripts were added.
7. don-petry

   closed this as completedin #1313on Jul 18, 2026
8. don-petry

   added a commit that references this issue on Jul 18, 2026

   feat: implement issue #1307 — [bmad-onboard S1] Onboard sre-lead pers…

   ...

   ec8a168