go-ldd v2.9.0: empathy test, decoder-ring refs, repo-idiom, when-in-Rome check

go-linter-driven-development/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "go-linter-driven-development",
-  "version": "2.8.0",
+  "version": "2.9.0",
   "description": "Rules-as-data linter-driven development workflow for Go: 12 single-source rule files, thin directional skills (design, TDD implementation, refactoring, testing, review, documentation), and a hunter/skeptic/critic review architecture with parallel single-rule reviewers, an over-abstraction skeptic, and a comment critic enforcing the Comment Value Toolbox",
   "author": {
     "name": "Dan Mordechay"

go-linter-driven-development/CHANGELOG.md

@@ -3,6 +3,54 @@
 All notable changes to the `go-linter-driven-development` plugin are documented here.
 Format based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); versions follow [Semantic Versioning](https://semver.org/).
 
+## [2.9.0] - 2026-07-23
+
+Lessons from a repo owner's hard review of a large generated PR: the reviewer
+could merge the code but "could not reasonably ask another human to maintain
+the result". Two themes: comments that need the code (or a decoder ring) to be
+understood, and a feature PR that silently adopted new engineering practices.
+
+### Added
+
+- **Empathy test — the plain-English test grows teeth** (R9 test 3, renamed
+  plain-English/empathy test). New persona anchor: write for a fresh graduate
+  whose first language may not be English. New failure modes: unexplained
+  acronyms and insider jargon ("DTO", "tristate") in comments AND in the symbol
+  names they document; and **self-standing** — a comment must be understandable
+  BEFORE reading the code ("if I need to read the code to understand the
+  comment, the comment adds negative value"); forward references to other
+  comments fail. The comment-critic now reads each comment before the
+  surrounding code so the self-standing check is built into its protocol.
+- **Decoder-ring references join the provenance anti-toolbox** (R9 floor +
+  toolbox catalog + critic): plan/decision/test-plan IDs ("T-04-02", "D-07"),
+  requirement tags ("REQ-SVC-01"), spec section refs ("spec §4") fail even
+  when they resolve inside a repo doc — the fact goes in the comment as plain
+  prose, the doc through one See-edge, the ID stays in the doc. Worked ❌/✅
+  example (`userResponse`) added to the catalog.
+- **Restated repo idiom is a floor failure mode** (R9 + catalog + critic): a
+  comment justifying a convention the repo applies everywhere (pointer field =
+  "omitted vs explicit zero") is noise even though it is technically a WHY —
+  the convention lives once at rung 2. The critic greps the repo for the same
+  pattern before crediting such a rationale.
+- **See-edge placement rule** (R9): the `See docs/<feature>.md` line is free
+  under the budget ONLY as its own trailing line; a doc reference braided into
+  the summary sentence is clutter, not an edge.
+- **When-in-Rome check** (@pre-commit-review step 1 + new 🟠 New Practice
+  category + new maxim): anything the diff introduces that the host repo does
+  not already use — test mechanisms (golden files), dependencies, tools, config
+  conventions, repo-level file edits bundled into a feature diff — is flagged
+  as an adoption decision that belongs to the repo owner; the fix is a
+  discussion or a separate PR, never silent inclusion. Detection is
+  comparative: grep the repo outside the diff for prior use.
+- **Two new maxims**: "Empathy is a core engineering value" (borrowed from a
+  reviewer's coding standards) under Clarity and knowledge; "When in Rome,
+  code as the Romans do" under Process and economics.
+- **Name-jargon note in the critic**: renames are not the critic's to order,
+  but when the empathy test fails because the jargon lives in the symbol name
+  itself ("DTO"), the verdict block carries a rename recommendation for the
+  caller to route. (Added after a live run where the critic rewrote a comment
+  but left the jargon name unremarked.)
+
 ## [2.8.0] - 2026-07-20
 
 ### Added

go-linter-driven-development/agents/comment-critic.md

@@ -33,13 +33,19 @@ comments. Directives (`//go:`, `//nolint`, `// Output:`) are not comments; skip
 them.
 
 **Critique protocol, per comment:**
-1. Classify the symbol's tier (helper / contract / crossroads) from its role in
+1. Read the comment BEFORE the surrounding code, and note whether you understood
+   it standing alone. This ordering is itself the self-standing half of test 3:
+   a comment you only understood after reading the code fails ("if I need to
+   read the code to understand the comment, the comment adds negative value").
+2. Classify the symbol's tier (helper / contract / crossroads) from its role in
    the code — Read the surrounding code, don't guess from the name.
-2. Run the three tests from the payload, in order: toolbox-value (floor per line,
-   then ceiling for the whole comment against the tier), budget, plain English.
-3. For any failure, decide the smallest verdict that fixes it: cut lines (TRIM),
+3. Run the three tests from the payload, in order: toolbox-value (floor per line,
+   then ceiling for the whole comment against the tier), budget, plain English +
+   self-standing (the empathy test — judge it for a fresh graduate whose first
+   language may not be English; unexplained acronyms and insider jargon fail).
+4. For any failure, decide the smallest verdict that fixes it: cut lines (TRIM),
    replace content (REWRITE), or remove entirely (DELETE).
-4. Every TRIM/REWRITE ships the proposed replacement text, and the proposal names
+5. Every TRIM/REWRITE ships the proposed replacement text, and the proposal names
    the toolbox item it delivers ("swap narrated implementation for the boundary
    contract this parsing constructor needs"). A bare "too long" is not a verdict.
 
@@ -48,9 +54,29 @@ citations, "the previous behavior" narration, and "matching what <old system>
 did" fail the floor even when the surrounding WHY is good — a reader five years
 out cares how the product behaves now, not which review round shaped it. Verdict
 TRIM (cut the provenance tail) or REWRITE (restate the history as present-tense
-rationale: "silently resolving by precedence would pick a mode the caller didn't
-ask for"). An incident/ticket reference survives only when it IS the rationale
-for a constraint.
+rationale: "silently picking one of the TLS options could apply a mode the
+caller did not ask for"). An incident/ticket reference survives only when it IS
+the rationale for a constraint.
+
+**Decoder-ring references are provenance in a different costume:**
+plan/decision/test-plan IDs ("T-04-02", "D-07"), requirement tags
+("REQ-SVC-01"), spec section refs ("spec §4"). They fail even when the token
+resolves inside a repo doc — a reader without the decoder ring gets nothing.
+REWRITE: the fact as plain prose, the doc via one trailing See-edge, the ID
+gone.
+
+**Jargon in the symbol name:** your verdicts are about comments, and a rename
+is not yours to order. But when the empathy test fails because the jargon
+lives in the symbol name itself ("DTO", "mgr"), say so — append a
+`note: symbol name carries the jargon — recommend rename (e.g. userDTO →
+userResponse)` line to the verdict block so the caller can route it.
+
+**Repo idiom is not a WHY:** before crediting a rationale that justifies a
+mechanical pattern (a pointer field for "omitted vs explicit zero", the
+standard error-wrapping style), grep the repo for the same pattern. If it
+appears across packages uncommented, this comment restates a repo-wide
+convention — verdict DELETE; the convention's home is the coding-standards doc,
+not a use site.
 
 **Boundary with R3:** an in-body comment that names what the next block does is an
 extraction candidate, not a rewrite candidate — verdict `DELETE → route R3`

go-linter-driven-development/maxims.md

@@ -190,6 +190,19 @@ imagined 30%. (R12's ceremony-copy inverse is one instance: don't "optimize" *or
 
 **Uncompiled** — no rule owns profile-first optimization discipline yet.
 
+### When in Rome, code as the Romans do
+— folk wisdom; sharpened by a repo owner's hard review of a large generated PR
+
+**Ask:** does this diff arrive in the host repo's existing style, or does it
+import mine? A new test mechanism, dependency, framework, or convention — however
+good — is an *adoption decision* that belongs to the repo owner, not a side
+effect of a feature PR. Strong opinions travel as a discussion or a separate PR,
+never as a bundled surprise. Reviewers forgive imperfect code in the house style
+far more readily than perfect code in a foreign one.
+
+**Compiled into:** @pre-commit-review's when-in-Rome check (step 1) and its
+🟠 New Practice finding category.
+
 ## Clarity and knowledge
 
 ### Clear is better than clever
@@ -200,6 +213,20 @@ future reader.
 
 **Compiled into:** `rules/R3-storifying.md` and the plugin's readability philosophy.
 
+### Empathy is a core engineering value
+— borrowed from a reviewer's coding standards ("coding in a way that facilitates
+maintenance by peers who may be less skilled or experienced")
+
+**Ask:** could a fresh graduate whose first language may not be English read this
+comment — before reading the code — and understand it? If you need to read the
+code to understand the comment, the comment adds negative value. Codenames,
+unexplained acronyms, and cross-reference webs serve the writer's bookkeeping,
+not the reader.
+
+**Compiled into:** `rules/R9-repo-brain.md`'s plain-English/empathy test (test 3
+of the three-test standard: persona, acronym ban, self-standing requirement) and
+the comment-critic's read-comment-first protocol.
+
 ### Once and only once
 — Kent Beck (and DRY: "every piece of knowledge has a single, unambiguous,
 authoritative representation" — Hunt & Thomas)

go-linter-driven-development/rules/R9-repo-brain.md

@@ -201,30 +201,48 @@ comment-critic agent enforces them adversarially after writing:
    - *Floor (delete test):* a prose line that delivers none of the toolbox values
      is trash — cut it. Named failure modes: restated identifiers, generic filler
      ("provides validation functionality"), narrated implementation, menu sections
-     filled without earning their place, and **provenance** — PR numbers, review
-     items, "the previous behavior" narration, "matching what <old system> did".
+     filled without earning their place, **restated repo idiom** — a comment
+     justifying a convention the repo already applies everywhere (a pointer field
+     meaning "omitted vs explicit zero", the standard error-wrapping style); the
+     convention is documented once at rung 2 (coding standards), never
+     re-explained at each use site — and **provenance and decoder-ring
+     references** — PR numbers, review items, plan/decision/test-plan IDs
+     ("T-04-02", "D-07"), requirement tags ("REQ-SVC-01"), spec section refs
+     ("spec §4"), "the previous behavior" narration, "matching what
+     <old system> did". A decoder-ring token fails even when it resolves inside a
+     repo doc: the reader gets the fact as plain prose and the doc through the
+     one See-edge — never through a code they must look up.
      The floor's lens is the 5-year reader test: a reader five years out cares how
      the product behaves NOW, never which PR or review round produced it. History
-     is rewritten as present-tense rationale ("silently resolving by precedence
-     would pick a TLS mode the caller didn't ask for"), and an incident/ticket
-     reference survives only when it IS the rationale for a constraint — never as
-     provenance.
+     is rewritten as present-tense rationale ("silently picking one of the TLS
+     options could apply a mode the caller did not ask for"), and an
+     incident/ticket reference survives only when it IS the rationale for a
+     constraint — never as provenance.
    - *Ceiling (smart choice):* the comment as a whole must carry the
      highest-value toolbox items for that symbol's tier within its budget. A
      crossroads whose five lines are all boundary trivia while the architectural
      WHY is missing fails, even though each line individually "adds something".
 2. **Budget test** — the comment fits its tier budget (accounting and tiers
    below).
-3. **Plain-English test** — plain English only: everyday words, short sentences,
-   one idea per sentence. Not all readers are native English speakers; a comment
-   that needs a dictionary fails even when true and within budget. Failure modes:
-   fancy vocabulary where a common word exists ("utilize" → "use", "leverages" →
-   "uses"), stacked clauses, academic phrasing.
+3. **Plain-English test (the empathy test)** — write for a fresh graduate whose
+   first language may not be English: everyday words, short sentences, one idea
+   per sentence. A comment that needs a dictionary fails even when true and
+   within budget. Failure modes: fancy vocabulary where a common word exists
+   ("utilize" → "use", "leverages" → "uses"), stacked clauses, academic phrasing,
+   and unexplained acronyms or insider jargon ("DTO", "tristate") — in the
+   comment AND in the symbol name it documents.
+   The test has a second half, **self-standing**: the comment must be
+   understandable BEFORE reading the code. If the reader must read the code — or
+   another comment ("see X's doc comment for why") — to understand this comment,
+   it has negative value. State the fact in place; forward references to other
+   comments fail.
 
 **Budget accounting** — prose lines count; these are free:
 
 - blank `//` separator lines
-- the `See docs/<feature>.md` network-edge line
+- the `See docs/<feature>.md` network-edge line — free ONLY as its own trailing
+  line; a doc reference woven into a prose sentence is not an edge, it is clutter
+  in that sentence's line count
 - short inline example lines, bounded at 2–4 lines — anything bigger belongs in an
   `Example_*` testable example
 
@@ -261,8 +279,10 @@ rule's (Q4 below — they must carry why/context, not restate the identifier).
 ### Edge conventions
 
 - **Code → docs** (rung 1 → rung 2): a literal relative path in the comment —
-  `See docs/retry-policy.md`. Paths to docs are fine; docs move rarely and Q2
-  verifies them mechanically.
+  `See docs/retry-policy.md` — always on its own trailing line, never braided
+  into the summary sentence (the first sentence stays clean: "Package accounts
+  registers the /accounts REST endpoints.", then the See-line). Paths to docs
+  are fine; docs move rarely and Q2 verifies them mechanically.
 - **Docs → code** (rung 2 → rungs 0–1): cite by **exported symbol** (`Policy`,
   `ParsePolicy`); a **package or directory path** (`retry/`) only when a location
   is genuinely needed; **file paths never, line numbers never** — they are the most

go-linter-driven-development/skills/documentation/SKILL.md

@@ -45,9 +45,14 @@ entry (worked examples in reference.md, "Bug Fix Documentation").
 doc that gets skipped.
 
 **Plain words over clever words**: everyday English, short sentences, one idea per
-sentence. Not all readers are native English speakers — a doc that needs a
-dictionary fails even when it is true (R9's plain-English test; applies to godocs
-and feature docs alike).
+sentence. Write for a fresh graduate whose first language may not be English — a
+doc that needs a dictionary fails even when it is true (R9's plain-English/empathy
+test; applies to godocs and feature docs alike).
+
+**Comments stand alone**: a reader must understand the comment BEFORE reading the
+code. If reading the code is required to understand the comment, the comment adds
+negative value — no decoder-ring IDs, no forward references to other comments
+(R9's empathy test, second half).
 </philosophy>
 
 <mode_selection>