This is the record of an abcd-managed repository

abcdBetav0.16.2 · 2026-11-06

Development record

a superseded Phase-0 pseudo-ADR sits in research/notes/ claiming ADR-01, outside every record gate

issueresolvedmajoriss-2608221457227162

a superseded Phase-0 pseudo-ADR sits in research/notes/ claiming ADR-01, outside every record gate

.abcd/development/research/notes/01-harness-interface.md is shaped like an accepted decision but filed as evidence, and the mismatch has let a superseded architecture sit in the record unchallenged. Four defects, one root cause.

  1. It presents as a decision in the evidence folder. Line 1 is # ADR-01: Harness Interface Design and it declares Status: Accepted (Phase 0 lock). The research/notes/README.md "What does NOT belong here" list opens with decisions, which belong in .abcd/development/decisions/adrs/.
  2. The id collides. The real adr-1 is decisions/adrs/0001-three-layer-mental-model.md, with id: adr-1 frontmatter. Two documents claim ADR-01.
  3. It carries no frontmatter, so it has no id, no status, and no superseded_by. record-lint's ADR machinery never sees it as a record, which is the mechanism behind the staleness: an "accepted lock" describing harness.py, pluggy, abc.ABC, and anthropic.types.Message passed through adr-21 (rebuild-in-go, 2026-07-06) untouched, because no gate was watching it.
  4. Two inbound references dangle. The brief, at 05-internals/03-configuration.md:403, gives the ADR naming convention as NNNN-<slug>.md (e.g. 0001-harness-interface.md), citing a filename that does not exist; itd-6 line 73 cites "ADR-02 (02-mcpbridge-implementation-contract.md)", which exists nowhere in the repo.

Why it matters: the document is not held to currency by adr-5, because it is not the brief. The defect is that it reads as normative while sitting outside every gate that governs normative records, so a reader reasonably takes a superseded Python design for current architecture. That misreading has already happened once in review.

Note on discoverability: the harness-name normalisation (PR #442) rewrote 22 occurrences in this file, so its git mtime now suggests active maintenance. The content is unchanged by that sweep and remains superseded.

Grounds

  • pursued: the note no longer presents as a decision; a record-lint run at e076c8e3 with cross_store_id_claim armed reports nothing on it, and a reintroduced ADR-01 title would show it wrong by tripping that rule

Frontmatter

idiss-2608221457227162
typeissue
lifecycleresolved
severitymajor

mentions8