AI Development Methodology

E06 — BMAD v6 landscape pass (context-handoff + review-triage conventions)

Pillar (primary): P9 — Self-improvement velocity Pillar (secondary): P1 — Doc completeness Status: done (2026-08-14) Phase: Phase 1 — Foundation Started: 2026-08-14 (chartered and executed same day, at maintainer direction) Closed: 2026-08-14 — all five items shipped in v1.25.0 Owner: maintainer + AI coding agent

Outcome (jobs-to-be-done)

When the maintainer reviews a peer methodology and finds mechanisms that close real gaps in ours, they want those findings converted into scoped, individually-shippable backlog items — rather than a loose notes file — so each idea either ships in a versioned methodology release or is explicitly deferred/rejected with reasoning, and nothing valuable evaporates.

This epic is the intake for the BMAD-METHOD v6.11.0 review of 2026-08-14 (full analysis: self-development/brief/03_competitive_landscape.md, local-only/gitignored). Five Tier-1 ideas are chartered as items; five Tier-2 ideas are parked in FUTURE.md.

Exit criteria (binary)

Each of BL-0015…BL-0019 is resolved: shipped in a versioned methodology release, moved to FUTURE.md, or rejected — with the decision and reasoning recorded in the item (and in the CHANGELOG for shipped ones). — All five shipped in v1.25.0; each carries a **Resolution:** note in ARCHIVE.md.
No shipped item duplicates a convention already present in methodology/ (the v1.20–v1.23 imports are the known overlap risk; each item’s draft names what it extends). — Cross-AI contradiction check ran explicitly against the two-stage review (v1.22), EARS and active-context (v1.21), protected regions (v1.23), and the lock rules; the one real duplication found (the ran-only test rule stated three times) was reduced to a single canonical statement.
Every shipped change is markdown-only: no scripts, no installers, no new runtime dependencies (trust posture unchanged). — Verified: the diff touches only .md files.
The competitive-landscape brief’s BMAD entry is marked reviewed/current at close (or annotated stale if BMAD v7 shipped meanwhile). — Entry dated 2026-08-14, pinned to BMAD commit c96b7d1 / v6.11.0, with a next-review trigger on BMAD v7.

KPIs

Import velocity: ≥1 item shipped per methodology release cycle while the epic is active (matching the v1.20–v1.23 one-pass-per-release rhythm). Not gating.
Durability: at the next semi-annual self-evaluation, shipped conventions are still present and referenced (not reverted or dead text). Forward-looking; not gating closure.

Out of scope

Personas, menus, party-mode. Rejected in the review — token overhead + tooling coupling; our challenge-prompt / cross-AI / decision-ownership mechanisms cover the goals.
Installers, web bundles, platform targeting. Contradicts the documented no-CLI decision (brief/07_tech.md).
Elicitation/brainstorming catalogs. Conflicts with the kickoff stance “ask only what you genuinely cannot guess.”
Any executable tooling (memlog-style derived artifacts, validators, snapshot renderers). The idea “deterministic jobs shouldn’t be judgment calls” stays as advice; scripts remain the adopter’s stack decision.
Public comparison content. The analysis stays in the gitignored brief per the v1.17.1 decision; nothing in this epic adds competitor comparisons to committed docs.

Tier note (authority)

Every item in this epic proposes changes to methodology/*.md ⇒ T2 under the tier matrix. The autonomous loop MUST NOT execute these edits (loop Constraint 1). The loop/agents may draft proposed wording in loop-notes/ or in the item body; the maintainer authors and ships each change as a normal versioned release. Template-only portions (e.g. BL-0015/BL-0016 touching templates/) are T1-eligible only if the maintainer explicitly splits them out.

Linked docs

Pillars: P9 — Self-improvement velocity (primary), P1 — Doc completeness (secondary)
Analysis source: self-development/brief/03_competitive_landscape.md (local-only, gitignored — ask the maintainer for the file; it is not in the public repo)
Methodology targets: 04_backlog_items.md, 06_working_principles.md, 07_definition_of_done.md, 10_testing_and_verification.md, 08_lessons_and_memory.md
Precedent: CHANGELOG v1.20.0–v1.23.0 (the prior landscape-informed passes and their format)

Item roster

See BACKLOG.md for active items (BL-0015…BL-0019), ARCHIVE.md for completed, FUTURE.md for deferred (BL-0020…BL-0024).

Open questions

Batching: BL-0017 and BL-0018 both touch 07 + 10 — ship as one release (one coherent “review rigor” pass) or two? Lean: one release, two CHANGELOG bullets.
Where does the Code Map convention live (BL-0015): inside 04_backlog_items.md as an item-body section, or as a template fragment? Lean: 04 defines it, templates carry a slot.
Frozen-intent marker syntax (BL-0016): HTML comment (), a blockquote badge, or a heading convention? Must stay greppable and render cleanly on GitHub. Decide during drafting.

Risks

Convention bloat. Five additions to already-long docs (04 is 894 lines) push against the ~1,050-line soft cap and against doc clarity (P2). Mitigation: each item names its target section and adds, not appends — prefer tightening existing sections over new ones; if 04 would cross the cap, coordinate with the E03 pattern (trim/split) first.
Import without fit. BMAD’s mechanics assume a single-pipeline orchestrator; ours assumes parallel peers. Mitigation: each item’s draft must restate the idea in our vocabulary (items/locks/DoD), not BMAD’s (specs/build steps) — the brief’s §6 table is the mapping.
Duplicating prior imports. v1.21 EARS and v1.22 two-stage review are adjacent to BL-0015/BL-0017. Mitigation: exit criterion 2; drafts must cross-link what they extend.
BMAD v7 ships mid-epic and invalidates source pointers. Mitigation: pointers in the brief are commit-pinned (c96b7d1); ideas stand on their own once restated in our vocabulary.

This site is open source. Improve this page.