@windyroad/itil 0.61.1 → 0.61.2

package/skills/work-problems/test/work-problems-step-6-5-fix-and-continue.bats

@@ -1,254 +0,0 @@
-#!/usr/bin/env bats
-
-# P140: /wr-itil:work-problems Step 6.5 Failure handling subsection must
-# document diagnose-then-classify routing — fix-and-continue for the
-# documented mechanically-fixable allow-list, halt for everything else.
-#
-# Prior behaviour was a uniform halt-on-CI-failure rule that converted
-# 1-line stale-grep-string updates and transient flakes into ~45min queue
-# stalls, regressing JTBD-006 "Progress the Backlog While I'm Away"
-# without any governance benefit. P140's Phase 1 amendment replaces that
-# uniform rule with a closed allow-list policy authorising silent
-# fix-and-continue per ADR-013 Rule 5, capped at 3 retries per iteration
-# before falling back to the halt branch.
-#
-# Doc-lint contract assertions per ADR-037 Permitted Exception
-# (contract-assertion class — same shape as the P130 / P126 / P135
-# sibling fixtures). The asserted prose IS the load-bearing policy
-# surface — re-reading the SKILL.md is the only way an AFK reader (and
-# the iteration subprocess) learns the fixable-class taxonomy and the
-# retry cap. Behavioural verification is impossible until Phase 2's
-# advisory classifier ships (deferred per the ticket Fix Strategy —
-# observe over 30 days).
-#
-# @problem P140
-# @adr ADR-013 (Rule 5 — policy-authorised silent action)
-# @adr ADR-014 (one-commit-per-iter; retries each ride their own commit)
-# @adr ADR-018 (inter-iteration release cadence; this refines its
-#       Failure handling clause)
-# @adr ADR-026 (agent output grounding — diagnostic preamble citation)
-# @adr ADR-037 (skill-testing strategy — contract-assertion class)
-# @adr ADR-042 (above-appetite branch — Rule 3 commit-gate-per-retry
-#       precedent composes with this fix-and-continue branch)
-# @adr ADR-044 (decision-delegation contract — framework-resolution
-#       boundary; closed allow-list extensions are deviation-candidates)
-# @jtbd JTBD-006 (Progress the Backlog While I'm Away — primary)
-# @jtbd JTBD-001 (Enforce Governance Without Slowing Down — composes;
-#       per-retry gates preserve governance)
-
-setup() {
-  REPO_ROOT="$(cd "$(dirname "$BATS_TEST_FILENAME")/../../../../.." && pwd)"
-  SKILL_MD="$REPO_ROOT/packages/itil/skills/work-problems/SKILL.md"
-}
-
-@test "work-problems P140: SKILL.md exists" {
-  [ -f "$SKILL_MD" ]
-}
-
-# ── Failure handling subsection identity ───────────────────────────────────
-
-@test "work-problems P140: Step 6.5 Failure handling subsection cites P140" {
-  # The amendment must self-identify so future readers tracing back from
-  # the ticket find the load-bearing prose without keyword-guessing.
-  run grep -nE 'Failure handling.*P140|P140.*Failure handling' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Diagnostic preamble (ADR-026 grounding) ────────────────────────────────
-
-@test "work-problems P140: Failure handling cites gh run view --log-failed as the diagnostic preamble" {
-  # ADR-026 grounding: the orchestrator MUST read the actual failure
-  # output before classifying. Without this, classification degrades to
-  # guess-from-context.
-  run grep -nE 'gh run view.*--log-failed' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Failure handling cites ADR-026 (grounding) on the diagnostic preamble" {
-  # The grounding requirement should cite ADR-026 explicitly so the
-  # connection is auditable.
-  run grep -nE 'ADR-026' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Fixable-in-iter allow-list (closed) ────────────────────────────────────
-
-@test "work-problems P140: Failure handling names P081-class stale-grep-string as a fixable class" {
-  run grep -nE 'P081-class stale-grep-string|stale-grep-string' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Failure handling names hook stub mismatch as a fixable class" {
-  run grep -niE 'hook stub mismatch' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Failure handling names test ID drift as a fixable class" {
-  run grep -niE 'test ID drift' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Failure handling names environmental flake as a fixable class" {
-  run grep -niE 'environmental flake' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: allow-list is framed as 'closed' (not extensible at agent discretion)" {
-  # JTBD review guard-rail: persona could misread "fix-and-continue" as
-  # "auto-fix anything" without the closed framing. Future agent edits
-  # must not drift the allow-list open without explicit user direction.
-  run grep -niE 'allow-list.*closed|closed.*allow-list' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: extending the allow-list is framed as a deviation-candidate per ADR-044" {
-  # ADR-044 framework-resolution boundary: the closed list IS the
-  # framework-resolved policy. Adding a class is a direction-setting
-  # decision, not a mechanical fix.
-  run grep -niE 'deviation-candidate.*ADR-044|ADR-044.*deviation' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: ambiguous classification defaults to halt (no diagnose-then-guess)" {
-  # JTBD review guard-rail (b): without this, the persona-misread risk
-  # of "auto-fix anything" re-enters via fuzzy classification.
-  run grep -niE 'Ambiguous classification defaults to halt|ambiguous.*halt' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Fix-and-continue branch ────────────────────────────────────────────────
-
-@test "work-problems P140: Failure handling documents a fix-and-continue branch" {
-  run grep -niE 'Fix-and-continue branch|fix-and-continue branch' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: each fix-and-continue retry rides standard ADR-014 commit gate flow (architect / JTBD / risk-scorer)" {
-  # Architect-flagged invariant: governance gates MUST run on every
-  # retry. The fix-and-continue branch does NOT bypass gates.
-  run grep -niE 'standard ADR-014 commit gate flow|ADR-014.*commit gate' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: ADR-042 Rule 3 commit-gate-per-retry precedent is cross-referenced" {
-  # ADR-042 already establishes that retries each ride their own
-  # commit through full gate flow. P140 composes with that precedent
-  # rather than inventing a new commit-cardinality rule.
-  run grep -niE 'ADR-042 Rule 3' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── 3-retry cap (per iteration) ────────────────────────────────────────────
-
-@test "work-problems P140: Failure handling caps fix-and-continue at 3 retries" {
-  run grep -niE '3-retry cap|3 retr|three retr' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: 3-retry cap is per-iteration, not per-failure-class" {
-  # Without this clarification, an agent could reset the counter on
-  # each new failure class and drain budget indefinitely.
-  run grep -niE 'per[- ]iteration, not per[- ]failure[- ]class|cap is per[- ]iteration' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Halt branch preserved ──────────────────────────────────────────────────
-
-@test "work-problems P140: Halt branch preserved for genuinely-unrecoverable failures" {
-  run grep -niE 'genuinely-unrecoverable|genuinely unrecoverable' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Halt branch enumerates auth failure / npm publish rejection / semantic test as unrecoverable" {
-  # The halt branch's allow-list mirror — naming the unrecoverable
-  # classes makes the boundary auditable.
-  run grep -niE 'auth failure|npm publish rejection|semantic test.*judgment' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Step 2.5b cross-reference preserved (P126) ─────────────────────────────
-
-@test "work-problems P140: Halt branch routes through Step 2.5b surfacing routine (P126 preserved)" {
-  # The halt branch's existing P126 cross-reference must survive the
-  # amendment — surfacing accumulated user-answerable skips before
-  # emitting the halt summary remains the contract.
-  run grep -nE 'Step 2\.5b cross-reference \(P126\)' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── ADR-013 Rule 5 policy-authorised silent action ─────────────────────────
-
-@test "work-problems P140: fix-and-continue branch is policy-authorised per ADR-013 Rule 5" {
-  # ADR-044's framework-mediated surface includes "policy-authorised
-  # silent proceed" — the closed allow-list IS the policy. Future
-  # readers must find the citation to confirm this is not an ad-hoc
-  # bypass of Rule 1.
-  run grep -nE 'ADR-013 Rule 5|Rule 5 policy-authorised|policy-authorised.*ADR-013' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Composition cross-references ───────────────────────────────────────────
-
-@test "work-problems P140: Failure handling cross-references P081 (stop-gap composition)" {
-  # P081 is the structural-tests-are-wasteful root cause. Most
-  # P081-class stale-grep-string failures are P081's territory.
-  # Fix-and-continue is the stop-gap; P081's full retrofit is the
-  # structural elimination.
-  run grep -nE 'P081' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Failure handling cross-references P135 (decision-delegation contract)" {
-  # P135 + ADR-044 frame the closed allow-list as the
-  # framework-resolved policy.
-  run grep -nE 'P135' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Failure handling cross-references P130 (orchestrator main-turn ask discipline)" {
-  # P130 ensures fix-and-continue does NOT introduce mid-iter asks —
-  # the closed allow-list resolves the decision per ADR-044's
-  # framework-resolution boundary.
-  run grep -nE 'P130' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Failure handling cross-references P132 (over-ask in interactive sessions)" {
-  # P140 is the inverse of P132 on the failure-handling surface — both
-  # arise from over-defensive uniform routing. Naming the symmetry
-  # protects against future drift.
-  run grep -nE 'P132' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Mid-loop ask discipline halt-point bullet narrowed ─────────────────────
-
-@test "work-problems P140: Step 6.5 CI-failure halt-point bullet narrows to outside-allow-list / cap-reached scope" {
-  # The Mid-loop ask discipline subsection enumerates Step 6.5 CI-
-  # failure as a halt point. After P140 the halt fires only on
-  # unrecoverable failures — the bullet must reflect that narrower
-  # scope, otherwise future readers conclude all CI failures still
-  # halt.
-  run grep -nE 'fixable-in-iter allow-list|3-retry cap reached|outside the.*allow-list' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-# ── Non-Interactive Decision Making table row ──────────────────────────────
-
-@test "work-problems P140: Decision Making table carries a CI-failure-during-Step-6.5-drain row" {
-  # The decision table is the AFK reader's quick-reference; without a
-  # row here the failure-handling refinement is buried 80 lines up in
-  # Step 6.5.
-  run grep -nE '\| CI failure during Step 6\.5 drain' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Decision Making table row cites the closed fixable-in-iter allow-list" {
-  run grep -nE 'closed fixable-in-iter allow-list' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}
-
-@test "work-problems P140: Decision Making table row cites the 3-retry cap" {
-  run grep -nE 'CI failure during Step 6\.5.*3-retry cap|3-retry cap.*CI failure' "$SKILL_MD"
-  [ "$status" -eq 0 ]
-}

package/skills/work-problems/test/work-problems-step-6-5-postrelease-kv-callback.bats

@@ -1,209 +0,0 @@
-#!/usr/bin/env bats
-# Step 6.5 Post-release K→V auto-transition callback (P228) — behavioural.
-#
-# Driver: ADR-022 prescribes K→V transition on release, but until P228
-# there was no auto-fire surface to back-fill the transition once a fix
-# ships. The post-release callback enumerator
-# (`packages/itil/lib/enumerate-postrelease-kv-candidates.sh`) is the
-# data source: it walks `docs/problems/known-error/*.md`, invokes the
-# derive-release-vehicle helper per ticket, and emits one
-# `KV_CANDIDATE: P<NNN> | <changeset>` line per ticket whose changeset
-# has shipped (derive exit 0).
-#
-# The 2026-06-08 P220 witness — `## Fix Released` populated but K→V
-# deferred citing a misapplied P143 amendment — is the empirical bug
-# this enumerator wires the surface to close.
-#
-# Cases covered (behavioural — exercise the helper, assert observable
-# output; structural ban per ADR-005 / P081):
-#   1. known-error/ dir absent           → KV_CANDIDATES_SUMMARY: total=0
-#   2. known-error/ dir empty            → KV_CANDIDATES_SUMMARY: total=0
-#   3. ticket present, derive exit 0     → KV_CANDIDATE emitted; total=1
-#   4. ticket present, derive exit 2     → skipped silently; total=0
-#      (no `.changeset/<name>.md` ref in body — legacy ticket)
-#   5. ticket present, derive exit 3     → skipped silently; total=0
-#      (changeset still in working tree — unreleased)
-#   6. mixed cohort (3 tickets: shipped + legacy + unreleased)
-#                                        → only the shipped one emitted
-#   7. README.md inside known-error/ is excluded from the enumeration
-#   8. unknown derive exit code          → stderr warning + skip; total=0
-#
-# Cross-references:
-#   @problem P228 (K→V auto-transition gap)
-#   @adr ADR-022 (Verifying lifecycle)
-#   @adr ADR-018 (release-cadence host of the callback)
-#   @adr ADR-005 (behavioural bats per P081)
-#   @jtbd JTBD-006 (Progress the Backlog While I'm Away — primary driver)
-
-setup() {
-  REPO_ROOT="$(cd "$(dirname "$BATS_TEST_FILENAME")/../../../../.." && pwd)"
-  HELPER="$REPO_ROOT/packages/itil/lib/enumerate-postrelease-kv-candidates.sh"
-
-  FIXTURE="$(mktemp -d)"
-  mkdir -p "$FIXTURE/docs/problems"
-
-  STUB_DIR="$(mktemp -d)"
-}
-
-teardown() {
-  rm -rf "$FIXTURE" "$STUB_DIR"
-}
-
-# Helper — write a stub derive-release-vehicle command that returns a
-# canned exit code (per ticket-id-to-exit-code map encoded in $STUB_MAP).
-# $STUB_MAP shape: `<NNN>:<exit>:<changeset>;<NNN>:<exit>:<changeset>;...`
-# The third field is consumed only when exit=0 (emitted in the RELEASE_VEHICLE
-# block the lib greps for changeset path).
-write_stub_derive() {
-  local stub="$STUB_DIR/wr-itil-derive-release-vehicle"
-  cat > "$stub" <<'STUB'
-#!/usr/bin/env bash
-# Stub derive-release-vehicle for the enumerator behavioural test. Reads
-# $STUB_MAP from the environment, returns canned exit codes per ticket.
-nnn="$1"
-nnn_norm="$(printf '%03d' "$((10#$nnn))")"
-IFS=';' read -ra entries <<< "${STUB_MAP:-}"
-for entry in "${entries[@]}"; do
-  [ -z "$entry" ] && continue
-  IFS=':' read -r id ex cs <<< "$entry"
-  id_norm="$(printf '%03d' "$((10#$id))")"
-  if [ "$id_norm" = "$nnn_norm" ]; then
-    if [ "$ex" -eq 0 ]; then
-      cat <<EOF
-RELEASE_VEHICLE:
-  changeset: $cs
-  version-packages-commit: deadbeefdeadbeefdeadbeefdeadbeefdeadbeef
-  pr: #999
-  merge-commit: deadbeefdeadbeefdeadbeefdeadbeefdeadbeef
-  release-date: 2026-06-08
-EOF
-    fi
-    exit "$ex"
-  fi
-done
-# Default: no entry for this ticket → exit 1 (not found)
-exit 1
-STUB
-  chmod +x "$stub"
-}
-
-# Helper — write a known-error ticket fixture (body content is opaque to
-# the enumerator; only the filename's leading NNN matters since the
-# derive helper is stubbed).
-write_ke_ticket() {
-  local nnn="$1"
-  local slug="$2"
-  mkdir -p "$FIXTURE/docs/problems/known-error"
-  cat > "$FIXTURE/docs/problems/known-error/${nnn}-${slug}.md" <<EOF
-# Problem ${nnn}: ${slug}
-
-**Status**: Known Error
-
-## Description
-
-Test fixture for the enumerator behavioural bats.
-EOF
-}
-
-# Source the lib once for all tests.
-load_lib() {
-  # shellcheck source=/dev/null
-  source "$HELPER"
-}
-
-@test "helper file exists and is sourceable" {
-  [ -f "$HELPER" ]
-  load_lib
-}
-
-@test "case 1: known-error/ dir absent → total=0" {
-  load_lib
-  rm -rf "$FIXTURE/docs/problems/known-error"
-  run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=0"* ]]
-  [[ "$output" != *"KV_CANDIDATE:"* ]]
-}
-
-@test "case 2: known-error/ dir empty → total=0" {
-  load_lib
-  mkdir -p "$FIXTURE/docs/problems/known-error"
-  run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=0"* ]]
-  [[ "$output" != *"KV_CANDIDATE:"* ]]
-}
-
-@test "case 3: ticket present + derive exit 0 → KV_CANDIDATE emitted, total=1" {
-  load_lib
-  write_stub_derive
-  write_ke_ticket "228" "adr-022-known-error-to-verifying-transition"
-  PATH="$STUB_DIR:$PATH" STUB_MAP="228:0:.changeset/p228-fix.md" \
-    run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" == *"KV_CANDIDATE: P228 | .changeset/p228-fix.md"* ]]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=1"* ]]
-}
-
-@test "case 4: ticket present + derive exit 2 (no vehicle ref) → skipped silently, total=0" {
-  load_lib
-  write_stub_derive
-  write_ke_ticket "100" "legacy-pre-p330-no-vehicle"
-  PATH="$STUB_DIR:$PATH" STUB_MAP="100:2:" \
-    run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" != *"KV_CANDIDATE:"* ]]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=0"* ]]
-}
-
-@test "case 5: ticket present + derive exit 3 (unreleased) → skipped silently, total=0" {
-  load_lib
-  write_stub_derive
-  write_ke_ticket "200" "changeset-still-in-tree"
-  PATH="$STUB_DIR:$PATH" STUB_MAP="200:3:" \
-    run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" != *"KV_CANDIDATE:"* ]]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=0"* ]]
-}
-
-@test "case 6: mixed cohort — only the shipped ticket is emitted, total=1" {
-  load_lib
-  write_stub_derive
-  write_ke_ticket "228" "shipped"
-  write_ke_ticket "100" "legacy-no-vehicle"
-  write_ke_ticket "200" "changeset-in-tree"
-  PATH="$STUB_DIR:$PATH" \
-    STUB_MAP="228:0:.changeset/p228-fix.md;100:2:;200:3:" \
-    run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" == *"KV_CANDIDATE: P228 | .changeset/p228-fix.md"* ]]
-  [[ "$output" != *"KV_CANDIDATE: P100"* ]]
-  [[ "$output" != *"KV_CANDIDATE: P200"* ]]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=1"* ]]
-}
-
-@test "case 7: README.md inside known-error/ is excluded from the enumeration" {
-  load_lib
-  write_stub_derive
-  mkdir -p "$FIXTURE/docs/problems/known-error"
-  printf '# Known Error Tickets\n' > "$FIXTURE/docs/problems/known-error/README.md"
-  PATH="$STUB_DIR:$PATH" STUB_MAP="" \
-    run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" != *"KV_CANDIDATE:"* ]]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=0"* ]]
-}
-
-@test "case 8: unknown derive exit code → stderr warning + skip, total=0" {
-  load_lib
-  write_stub_derive
-  write_ke_ticket "999" "weird-exit"
-  PATH="$STUB_DIR:$PATH" STUB_MAP="999:42:" \
-    run enumerate_postrelease_kv_candidates "$FIXTURE/docs/problems"
-  [ "$status" -eq 0 ]
-  [[ "$output" != *"KV_CANDIDATE:"* ]]
-  [[ "$output" == *"KV_CANDIDATES_SUMMARY: total=0"* ]]
-  # Stderr warning is captured into $output when bats merges stderr; the
-  # contract here is "skip silently AND don't lose audit signal".
-}

package/skills/work-problems/test/work-problems-stop-condition-questions.bats

@@ -1,85 +0,0 @@
-#!/usr/bin/env bats
-# Doc-lint guard: work-problems SKILL.md must surface outstanding design
-# questions before emitting `ALL_DONE` at stop-condition #2. In AFK mode
-# (the persona default per JTBD-006) the questions are listed in the
-# post-stop summary as an "Outstanding Design Questions" table. In
-# interactive invocations the skill may batch up to 4 questions through
-# a single `AskUserQuestion` call per ADR-013 Rule 1.
-#
-# Structural assertion — Permitted Exception to the source-grep ban
-# (ADR-005 / P011). These tests assert that the skill specification
-# document contains the stop-condition #2 question-surfacing step and its
-# supporting classifier taxonomy.
-#
-# Cross-reference:
-#   P053: docs/problems/053-work-problems-does-not-surface-outstanding-design-questions-at-stop.open.md
-#   ADR-013 Rule 1 / Rule 6: docs/decisions/013-structured-user-interaction-for-governance-decisions.proposed.md
-#   ADR-018: docs/decisions/018-inter-iteration-release-cadence-for-afk-loops.proposed.md
-#   @jtbd JTBD-006 (Progress the Backlog While I'm Away)
-#   @jtbd JTBD-001 (Enforce Governance Without Slowing Down)
-
-setup() {
-  SKILL_DIR="$(cd "$(dirname "$BATS_TEST_FILENAME")/.." && pwd)"
-  SKILL_FILE="${SKILL_DIR}/SKILL.md"
-}
-
-@test "SKILL.md exists (P053 precondition)" {
-  [ -f "$SKILL_FILE" ]
-}
-
-@test "SKILL.md stop-condition #2 has a pre-terminal question-surfacing step (P053)" {
-  # P053 fix: between detecting stop-condition #2 and emitting `ALL_DONE`,
-  # the skill must surface outstanding design questions. Accept a dedicated
-  # "Step 2.5" heading OR equivalent wording that names the pre-terminal
-  # question-batching step.
-  run grep -inE "Step 2\.5|pre-terminal question|question-batching|outstanding design question|surface (the )?(outstanding )?questions? before ALL_DONE" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-}
-
-@test "SKILL.md stop-condition #2 cites the 4-question AskUserQuestion cap (P053 + ADR-013)" {
-  # ADR-013 Rule 1 routes governance decisions through AskUserQuestion.
-  # AskUserQuestion's documented per-call limit is 4 options. The
-  # Step 2.5 block must name this cap so implementers do not silently
-  # overflow.
-  run grep -inE "cap at 4|up to 4 question|4 options? per AskUserQuestion|AskUserQuestion'?s? 4-(question|option) (limit|cap)|4-question cap|four questions? (per|at) a time" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-}
-
-@test "SKILL.md classifier records a skip-reason taxonomy (P053)" {
-  # Step 4 classifier must distinguish user-answerable skips from
-  # architect-design and upstream-blocked skips so Step 2.5 can select
-  # the user-answerable subset. Accept either an explicit three-bucket
-  # taxonomy OR the individual category names appearing near the
-  # classifier.
-  run grep -inE "user-answerable|user answerable" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-  run grep -inE "architect-design|architect design (judgment|required)" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-  run grep -inE "upstream-blocked|upstream blocked|upstream dependency" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-}
-
-@test "SKILL.md non-interactive fallback emits an Outstanding Design Questions table (P053 + ADR-013 Rule 6)" {
-  # Rule 6: when AskUserQuestion is unavailable (AFK default per
-  # JTBD-006), the skill must record the questions in the post-stop
-  # summary rather than asking. Accept either literal "Outstanding
-  # Design Questions" heading OR a named table with Ticket/Question
-  # columns that carries the same semantics.
-  run grep -inE "Outstanding Design Questions|outstanding.design.questions" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-}
-
-@test "SKILL.md stop-condition #2 cites ADR-013 Rule 6 fallback path (P053)" {
-  # The fallback must be traceable to the ADR it honours. Require an
-  # explicit Rule 6 citation in the stop-condition block.
-  run grep -inE "ADR-013.*Rule 6|Rule 6.*ADR-013" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-}
-
-@test "SKILL.md Final Summary template includes the Outstanding Design Questions section when relevant (P053)" {
-  # The Output Format block must show reviewers what the summary looks
-  # like with outstanding questions attached — not just describe it in
-  # prose elsewhere.
-  run grep -n "### Outstanding Design Questions\|## Outstanding Design Questions" "$SKILL_FILE"
-  [ "$status" -eq 0 ]
-}