Last refreshed: 2026-10-18 (BL-0069 listing housekeeping; 8 done, 2 active + 1 standing, 0 planned, 1 parked.)
| Epic | Title | Pillar (primary + secondary) | Status | Phase | Items (open/done) | Next milestone |
|---|---|---|---|---|---|---|
| E00 | Intake — real work not worth a charter | P9 Self-improvement velocity | standing (2026-09-05) | all | 2 / 6 | Never closes. The dogfood run for “focused mode”: same item format, same gates, no epic above it. Exempt from the WIP cap by declared deviation — the cap limits concurrent chartered effort and intake is not chartered effort. Its first three closed items were all work with no epic to belong to — including both sweep decisions that had been recorded and left unexecuted, and the audit brief that unblocks E10’s closing gate. |
| E02 | First semi-annual methodology self-evaluation pass | P3 Doc currency + P2 Doc clarity | done (2026-05-25) | Phase 1 | 0 / 5 | First epic closed; cycle validated; next pass 2026-11-25. |
| E01 | Examples folder | P1 Doc completeness + P6 Example richness | done (2026-05-25) | Phase 1 | 0 / 5 | Shipped in v1.15.0 (examples/ folder with tinker fictional project). |
| E05 | CHEATSHEET.md (one-page reference) | P1 Doc completeness | done (2026-05-25) | Phase 1 | — / 1 | Shipped in v1.15.0 (CHEATSHEET.md at repo root, 97 lines at close — the “~80” recorded here until 2026-08-19 was never accurate). Drifted to 144 and sat over E05’s 100-line hard exit criterion from v1.20.0 until the v1.31.0 trim. It is 98 lines today — two lines of margin after the v1.34.0 trim, and nothing enforces the cap. BL-0049 shipped the budget as a written, reproducible command rather than a check; the committed checker was declined. |
| E03 | Trim or split 09_git_workflow.md |
P2 Doc clarity | done (2026-08-14) | Phase 1 | 0 / 4 | Trim chosen over split; 1,026 → 798 lines (−22%), 24 → 20 sections, zero content lost. Shipped in v1.27.0. |
| E04 | Native templates for Cursor / Aider / Continue.dev | P4 Tool compatibility | parked — will not resume (2026-08-14) | Phase 1 | 0 / 0 | Dropped by maintainer decision: AGENTS.md + adaptation is the permanent answer. Charter preserved with the reasoning. Reopens only if an adopter reports adaptation actually failing. |
| E07 | Agentic-workflow landscape pass (context integrity + spec-time verification) | P9 Self-improvement velocity + P1 Doc completeness | done (2026-08-14) | Phase 1 | 0 / 5 | Chartered and closed same day at maintainer direction; 5 items shipped in v1.29.0. Personas and severity tiers rejected in review. |
| E06 | BMAD v6 landscape pass (context-handoff + review-triage conventions) | P9 Self-improvement velocity + P1 Doc completeness | done (2026-08-14) | Phase 1 | 0 / 6 | Chartered and closed same day at maintainer direction; 5 chartered items shipped in v1.25.0, plus BL-0021 promoted from FUTURE.md post-closure and shipped in v1.26.0. 4 Tier-2 ideas remain in FUTURE.md. |
| E08 | Role briefs per phase, plus a second pass over the E07 source | P1 Doc completeness + P9 Self-improvement velocity | done (2026-08-19) | Phase 1 | 0 / 4 | Chartered and closed same day at maintainer direction; 4 items shipped in v1.30.0. Personas rejected a third time; three further findings held with reasoning in the charter. |
| E09 | External landscape pass over six repos (skills-spec conformance + item-level clarification marker) | P4 Tool compatibility + P1 Doc completeness | done (2026-08-19) | Phase 1 | 0 / 3 | Chartered and closed same day at maintainer direction; three items shipped in v1.31.0. Four of six sources rejected outright. Cross-AI findings-verification waived by maintainer decision — the first landscape pass to close without it. |
| E10 | External baseline audit remediation | P2 Doc clarity + P8 Maintenance sustainability | active (2026-08-20) | Phase 1 | 0 / 14 | Chartered 2026-08-20 against an external cold audit of be93a05 that returned “Not sound for stated use”. Ten of eleven findings still stood at v1.31.0. Every Critical and High one is closed and staged for a single v1.32.0: the skill parses, the trunk exception is gone, the Test enum has one definition, pass is reserved for the required level, destructive operations split into two disjoint classes, trust follows provenance, and the lock’s authority claim matches what git enforces. All fourteen items closed — every Critical, High and Medium finding. Also shipped: SHA-pinned actions with the supply-chain claim narrowed, this repo’s first root instruction file, an adoption profile, the surface map, the release-evidence commands, and a sweep finding that only 6 of 16 conventions added v1.25.0–v1.31.0 have ever been exercised. Stays active until the release lands and a fresh cold re-audit clears it — a session cannot audit its own work. A committed checker and any new CI were declined — F-08 closes as a convention, not a control. |
| E11 | Assurance and multidisciplinary software squads | P1 Doc completeness + P2 Doc clarity | active (2026-09-15) | Phase 1 | 3 / 0 | v1.35.0 draft and local checks complete; independent verification and maintainer acceptance remain closing gates. |
Counts: 2 active (E10, E11) + 1 standing (E00, WIP-exempt), 0 planned, 8 done (E01, E02, E03, E05, E06, E07, E08, E09), 1 parked (E04 — will not resume). E00 does not close and does not consume a WIP slot; the exemption is declared in its charter rather than read into the rule.
WIP cap = 2 active epics (raised from 1 on E02 close, 2026-05-25). Reasoning: E02 closed cleanly end-to-end through the autonomous loop with the tier matrix in effect; the loop demonstrated discipline (escalate-on-doubt fired correctly; diff-verification caught real issues; maintainer-merge gate preserved). With one successful epic close on record, the cap can rise to 2 without inflating risk. The cap may rise to 3 (methodology default) after the second epic closes and the loop has demonstrated discipline across two distinct epic shapes.
E00-intake is exempt from the cap. It is standing, not chartered — it never closes, so on a literal reading it would occupy one of two slots permanently and leave a single slot for all real work. The cap limits concurrent chartered effort; intake is not chartered effort. Recorded as a declared deviation rather than a re-reading of the rule, and it is a design finding the dogfood run produced in its first hour — the kind that would have gone unnoticed had focused mode been published first and used later.
Inverse view: which epics touch each pillar.
| Pillar | Active epics | Planned epics | Coverage status |
|---|---|---|---|
| P1 Doc completeness | E11 (primary) | — | E01 + E05 (primary) done in v1.15.0; E06 (secondary) done in v1.25.0; E08 (primary) done in v1.30.0 — role briefs close the paste-able-prompt gap; E09 adds the item-level clarification marker (unreleased) |
| P2 Doc clarity | E10 (primary), E11 (secondary) | — | E02 secondary (done v1.14.0); E03 primary (done v1.27.0 — 09 trimmed to 798 lines). E10 is the first clarity work driven by an outside reader rather than by the maintainer’s own re-read, and it targets a failure the pillar had not named: the authoritative statement is usually right, and the surface shaped for copying is the one that is wrong |
| P3 Doc currency | — | — | E02 (primary) done 2026-05-25; next semi-annual pass due 2026-11-25 |
| P4 Tool compatibility | E09 (primary) | — | E04 (primary) parked 2026-08-14 — AGENTS.md + adaptation is the accepted answer, so the pillar no longer waits on native templates. E09 revives the pillar from the other end: rather than shipping per-tool templates, state which open format the skill already conforms to (unreleased) |
| P5 Adopter discoverability | E09 (tertiary) | — | Dormant (Phase 2 pillar) — E09’s conformance line is the first change aimed at it since the campaign route was closed on 2026-08-19 |
| P6 Example richness | — | — | E01 (secondary) done in v1.15.0 (examples/ folder shipped) |
| P7 Community feedback loop | — | — | Dormant (Phase 2/3 pillar) |
| P8 Maintenance sustainability | E10 (secondary) | — | Partly awake. The audit named the missing integrity checker as the root cause of the recurring count drift; a committed checker was declined on 2026-08-20 because this repo ships no runnable elements. E10 writes the evidence commands down instead, which is weaker on purpose and recorded as such. The pillar’s real work is still ahead |
| P9 Self-improvement velocity | E09 (BL-0036) | — | Three landscape-import epics closed: E06 (BMAD, v1.25.0 to v1.26.0), E07 (agentic workflow, v1.29.0), and E08 (secondary — a re-read of E07’s own source, v1.30.0). E09 is the fourth pass and the first to return mostly rejections — four of six sources yielded nothing new, which is itself the useful result |
Observation: Until 2026-08-20 every pillar with a chartered epic had had it closed, both WIP slots were free, and nothing was queued. E10 changes the shape of the observation rather than just the counts. Five of the eight closures (E03, E06, E07, E08, E09) ran to completion inside a single maintainer-directed session rather than a sustained active period, so the cap has still never been contended and raising it to 3 would be premature on that evidence. But the more useful pattern is the one the audit exposed: eleven-plus conventions shipped across four landscape passes, and an outside reader found ten live defects in the rules those passes were adding to. Same-day charter-to-close was not the cause, and speed is not the finding — every one of the ten was a contradiction between two copies of a rule, or a claim published without being checked. E10 is the first epic here that is expected to take more than a session, and the first whose exit criterion is somebody else’s verdict.
Two decisions block work, and both are on E10. They are carried as Needs clarification markers on their items, which means the items cannot reach ready until they are answered — the first real use of the convention v1.31.0 shipped.
Test enum (T2). Whether the canonical set is the eight-value enum, and whether a no-subset rule binds: a surface reproduces all eight, or carries none and links. Blocks BL-0043 and through it BL-0044.One judgement still carries forward from E09. It was the first landscape pass to close without cross-AI findings-verification — waived by maintainer decision. E06’s run of that gate returned 16 PASS / 2 FAIL with both failures real, so the waiver traded a step that has demonstrably caught defects for speed. It is on the record for the evaluation due 2026-11-25, and E10’s charter makes the gate non-waivable for that epic — repairing a class of “claims asserted rather than checked” while asserting the repair would be the same defect one level up.
The binding constraint has moved off the backlog entirely: it is the Phase 1 → Phase 2 transition (closed-beta readiness per methodology/12). That needs ≥ 2 external adopters recruited and structured feedback collected. The active-campaign route was closed on 2026-08-19: the maintainer deleted the staged distribution drafts on the position that a good project sells itself, so the passive channels already in place (GitHub topics, Pages, the awesome-list listings recorded in P5) are now the whole of the discovery surface. HUMAN_NEEDED.md is consequently empty. FEEDBACK.md becomes load-bearing the moment it lands.
If work is wanted before then, the cheapest source is still the four Tier-2 ideas parked in E06’s FUTURE.md — boundaries triad, epic context digest, brownfield discovery pass, human-review walkthrough. (The memory admission test was promoted from that list and shipped in v1.26.0.) Promoting any means re-opening E06 or chartering a successor.
Worth noticing: an empty backlog is a signal, not an achievement. It usually means either the project is genuinely between phases — which is the case here — or that nobody is filing what they notice. The next self-evaluation should check which. It should also check the opposite risk, which v1.25.0 through v1.30.0 made live: that work gets chartered because it is available to do rather than because anything demanded it. Three landscape passes and eleven-plus conventions have shipped without a single external adopter exercising them. E09 is the fourth pass and was scoped against that fact rather than around it: six sources reviewed, four rejected outright, one convention added, and its primary item aimed at adopter discoverability instead of at the rulebook. Whether that restraint was real or just better-narrated is a fair question for the next evaluation to put to it.
rejected epic state, so a decided-against epic is parked with an explicit will-not-resume marker and the reasoning in its charter (see E04).methodology/, templates/ or a root policy file, which makes them T2 or T3 under the tier matrix — maintainer-authored, loop-disabled. Two also carry unresolved Needs clarification markers, which independently blocks ready. The loop should halt and surface rather than promoting an epic itself (epic promotion is a maintainer decision).Update this rollup whenever:
BACKLOG.md (update the items column).Per the methodology, the rollup is a navigation aid — it should always reflect the current state of the underlying epic folders. Stale rollup = lost trust.