Status: Accepted
Date: 2026-07-06

Context

The .context/ frontmatter contract used mixed-case enum values: type and
status were lowercase (plan, draft), severity and value were lowercase
(high, critical), while effort was already upper (S/M/L/TBD). ADR-049
codified value with lowercase enums, consistent with the fields around it. The
casing was inconsistent across the contract and we chose to standardise it.

This ADR amends the enum casing established across ADR-046 (severity,
known-issue), the earlier effort/status/type conventions, and ADR-049
(value). It does not revisit any of their substantive decisions — the
three-axes model, the value rubric, the read protocol, and per-type requiredness
from ADR-049 all stand unchanged. Only the surface form of the values changes.

Decision

1. All .context/ frontmatter enum values are UPPER_CASE:

   • type: PLAN | FINDING | ANALYSIS | INSTRUCTION | AUDIT | KNOWN_ISSUE
   • status: DRAFT | ACTIVE | DONE | SUPERSEDED
   • severity: CRITICAL | HIGH | MEDIUM | LOW
   • value: HIGH | MEDIUM | LOW
   • effort was already S | M | L | TBD — unchanged.

2. known-issue becomes KNOWN_ISSUE (SCREAMING_SNAKE_CASE). The directory
   stays known-issues/; the generator maps the KNOWN_ISSUE type value to that
   directory group. The enum value and directory name intentionally differ in case
   and separator.

3. ADR status vocabulary is out of scope and stays lowercase. ADRs use
   proposed | accepted | deprecated | superseded, a separate vocabulary
   with its own tooling (merge-status-sync.sh, the ADR schema). It is not changed
   here. Where a script handles both (e.g. merge-status-sync-lib.sh), only the
   plan-status branch was uppercased; the ADR-status branch was left as-is.

4. The migration is mechanical and total. The schema enums, the validator’s
   per-type conditionals, the index generator’s type_group_key and
   severity_rank, check-plan-drift.sh, the plan-status branch of
   merge-status-sync.sh, the Go remediation-plan generator, every existing
   .context/**/*.md frontmatter value, the authoring skills and their templates,
   and the fixture tests were all converted in one PR.

Consequences

• Easier: one casing convention across the whole frontmatter contract; enum
  values are visually distinct from surrounding prose and directory names.
• Breaking, one-time: every .context/ file changed, so the migration had to
  land atomically with the validator, generator, and Go changes, or validation
  would fail mid-migration. Done as a single follow-up PR after the value-signal
  work (ADR-049) had merged, to keep that PR focused and green.
• Case-sensitive going forward: validate-context-frontmatter.sh rejects
  lowercase values (its enum checks derive from the schema). A file authored with
  status: active now fails validation with a clear message.
• Existing ADRs are unchanged. Per the ADR immutability rule, historical ADRs
  (including ADR-049 and ADR-046) keep their original lowercase examples as
  point-in-time records; this ADR is the forward-looking casing authority.
• Not addressed: ADR-status casing (deliberately out of scope) and the
  separate observation that Go-generated remediation plans do not yet carry a
  value field (tracked independently, not a casing concern).

Amendment (2026-07-06): effort deliberately stays S/M/L/TBD

Decision 1 recorded effort as “unchanged” almost in passing. This amendment
promotes that to an explicit, reasoned decision so it is not re-opened:

effort intentionally keeps its single-letter T-shirt sizes (S/M/L/TBD)
and is the one enum exempt from the full-word UPPER_CASE convention.

A proposed migration to full words (SMALL/MEDIUM/LARGE) was raised as the
effort-single-letter-values known-issue and taken through a 3-reviewer
plan-review (2026-07-06). All three reviewers recommended against it:

• No benefit: effort already parses and sorts correctly; nothing downstream
  depends on its spelling. Graded value: LOW.
• Net-negative: full words make MEDIUM a legal value for effort, value,
  and severity simultaneously — an ambiguity that cannot occur while effort
  stays single-letter, since S/M/L share no token with the word-based enums.
  The migration would trade one cosmetic inconsistency for a subtler collision.
• Disproportionate blast radius: ~73 occurrences across ~28 files, including an
  independent aggregationEffort() in reporter/aggregation.go that the test suite
  would not catch — a silent split-brain risk far exceeding the cosmetic gain.

The effort-single-letter-values known-issue is therefore closed SUPERSEDED by
this amendment. Single-letter effort is the deliberate, documented exception to
the UPPER_CASE convention, retained specifically to keep the three magnitude axes
(effort / value / severity) unambiguous at a glance.