The three opportunities the external audit raised alongside its findings. All are real and none is
chartered: each is either strictly larger than the finding it would strengthen, or blocked on something
this epic cannot supply. Deferred, not rejected — outright rejections live in the
charter’s “Out of scope”. Promote by moving an entry into BACKLOG.md
unchanged; this backlog uses the repo-wide monotonic BL-#### space for future items too, per
backlog/README.md, so promotion needs no renumbering.
All three entries are methodology/- or process-touching ⇒ T2 (maintainer-authored) under the tier
matrix.
Small synthetic repositories exercising the failure modes this epic fixes on paper: two agents acquiring
the same item from separate feature branches; a brownfield partial adoption; a wrong Code Map; a changed
AGENTS.md arriving in an untrusted PR; an L4-required feature sitting at L3. Run the same tasks through
supported agent tools and score the actions against a shared expected trace.
Source: audit opportunity O-01.
Target: a new fixtures directory, plus a scoring note per run.
Why it is deferred rather than chartered: every item in this epic is a documentation fix whose success criterion is “one governing statement per rule”. Fixtures answer a different and harder question — whether a cold agent, reading the fixed docs, actually behaves correctly. That is worth knowing and it is a larger piece of work than the repairs themselves. Chartering it now would also test text that is still changing.
Symptom that would justify promoting it: the re-audit passes, and a cold agent still picks the wrong reading of a rule this epic reconciled. That is the evidence that reconciliation was necessary but not sufficient.
Keep lifecycle policy, autonomy classes and authority in small versioned data files; generate the tables, checklists and skill excerpts from them; emit a release-evidence manifest carrying the tree SHA, checker version, counts and tag intent.
Source: audit opportunity O-02.
Target: would supersede BL-0048’s register and part of BL-0049’s checker.
Why it is deferred rather than chartered: this is strictly stronger than declare-then-check, and strictly more machinery. Generation removes the possibility of drift where checking only removes the persistence of it — but it also means the fourteen docs stop being hand-authored markdown, which is a different project. The cheaper control ships first and earns the harder one.
Symptom that would justify promoting it: the checker starts finding real drift repeatedly, which would mean the register is being ignored rather than followed.
Recruit at least two independent adopters in different repository shapes, record their deviations and failure rates, and have a reviewer who did not author the methodology score the transfer outcomes.
Source: audit opportunity O-03.
Target: no file — this is evidence collection, not a doc change.
Why it is deferred rather than chartered: it is already the Phase 1 → Phase 2 blocker recorded in
EPICS.md, and the active-campaign route was closed by maintainer decision on
2026-08-19 on the position that a good project sells itself. Recording it here keeps the audit’s decision
list fully accounted for; it does not duplicate a known blocker into a second place where it can go stale.
Note on where it does not belong: this is not a HUMAN_NEEDED.md entry. That file holds items
blocked on a human-only action. There is no blocked item here — only an absent adopter, which no action
in this repo produces.
Symptom that would justify promoting it: an adopter appears. Until then the constraint is external and the correct state is to wait, not to plan.