@kontourai/flow-agents 3.6.0 → 3.8.0

package/evals/integration/test_goal_fit_hook.sh

@@ -2242,12 +2242,12 @@ node - "$ROOT/scripts/hooks/stop-goal-fit.js" <<'NODEEOF' 2>"$TMPDIR_EVAL/hardbl
 const fs = require('fs');
 const file = process.argv[2];
 let src = fs.readFileSync(file, 'utf8');
-const needle = "const HARD_BLOCK = /contradicts evidence\\.json|caught false-completion|evidence verdict:|evidence check .+ status:|critique status|critique open|required sidecar is missing|command-log integrity check FAILED|gate misconfiguration:|exit-code-laundered|NOT_VERIFIED \\(ambiguous\\)/;";
+const needle = "const HARD_BLOCK = /contradicts evidence\\.json|caught false-completion|evidence verdict:|evidence check .+ status:|critique status|critique open|required sidecar is missing|command-log integrity check FAILED|gate misconfiguration:|exit-code-laundered|NOT_VERIFIED \\(ambiguous\\)|canonical Flow (?:run remains active|state is unsafe or malformed)/;";
 if (!src.includes(needle)) {
   process.stderr.write('mutation: HARD_BLOCK NOT_VERIFIED (ambiguous) pattern not found — source pattern drifted, cannot mutation-test\n');
   process.exit(1);
 }
-const mutated = "const HARD_BLOCK = /contradicts evidence\\.json|caught false-completion|evidence verdict:|evidence check .+ status:|critique status|critique open|required sidecar is missing|command-log integrity check FAILED|gate misconfiguration:|exit-code-laundered/;";
+const mutated = "const HARD_BLOCK = /contradicts evidence\\.json|caught false-completion|evidence verdict:|evidence check .+ status:|critique status|critique open|required sidecar is missing|command-log integrity check FAILED|gate misconfiguration:|exit-code-laundered|canonical Flow (?:run remains active|state is unsafe or malformed)/;";
 src = src.split(needle).join(mutated);
 fs.writeFileSync(file, src);
 NODEEOF
@@ -2712,6 +2712,64 @@ else
   _fail "mutation-test cleanup re-check REGRESSION: restored stop-goal-fit.js no longer confirms a clean end-to-end pass for the dialect fixture: $(cat "$TMPDIR_EVAL/dialect-restore-e2e.out" "$TMPDIR_EVAL/dialect-restore-e2e.err")"
 fi
 
+# Canonical Flow is authoritative for an active scoped session. Parent-directory
+# symlinks and malformed run state must therefore fail closed instead of being
+# silently treated as a missing optional run.
+CANONICAL_UNSAFE_REPO="$TMPDIR_EVAL/canonical-unsafe-repo"
+CANONICAL_UNSAFE_SESSION="$CANONICAL_UNSAFE_REPO/.kontourai/flow-agents/canonical-unsafe"
+mkdir -p "$CANONICAL_UNSAFE_SESSION"
+printf '# Test Repo\n' > "$CANONICAL_UNSAFE_REPO/AGENTS.md"
+cat > "$CANONICAL_UNSAFE_SESSION/canonical-unsafe--deliver.md" <<'MARKDOWN'
+# Canonical unsafe fixture
+
+status: executing
+type: deliver
+MARKDOWN
+cat > "$CANONICAL_UNSAFE_SESSION/state.json" <<'JSON'
+{
+  "task_slug": "canonical-unsafe",
+  "status": "executing",
+  "phase": "execution",
+  "flow_run": { "status": "active", "current_step": "verify" }
+}
+JSON
+cat > "$CANONICAL_UNSAFE_REPO/.kontourai/flow-agents/current.json" <<'JSON'
+{ "active_slug": "canonical-unsafe" }
+JSON
+CANONICAL_OUTSIDE="$TMPDIR_EVAL/canonical-outside"
+mkdir -p "$CANONICAL_OUTSIDE"
+ln -s "$CANONICAL_OUTSIDE" "$CANONICAL_UNSAFE_REPO/.kontourai/flow"
+if FLOW_AGENTS_GOAL_FIT_STRICT=true node "$ROOT/scripts/hooks/stop-goal-fit.js" >"$TMPDIR_EVAL/canonical-symlink.out" 2>"$TMPDIR_EVAL/canonical-symlink.err" <<JSON
+{"hook_event_name":"Stop","cwd":"$CANONICAL_UNSAFE_REPO"}
+JSON
+then
+  _fail "strict goal-fit hook allowed an active scoped session with a symlinked canonical Flow parent"
+elif [[ "$?" -eq 2 ]] && grep -q 'canonical Flow state is unsafe or malformed' "$TMPDIR_EVAL/canonical-symlink.err"; then
+  _pass "strict goal-fit hook rejects symlinked canonical Flow parent components for an active scoped session"
+else
+  _fail "strict goal-fit hook did not report the unsafe canonical Flow parent: $(cat "$TMPDIR_EVAL/canonical-symlink.err")"
+fi
+
+rm "$CANONICAL_UNSAFE_REPO/.kontourai/flow"
+mkdir -p "$CANONICAL_UNSAFE_REPO/.kontourai/flow/runs/canonical-unsafe"
+printf 'not-json\n' > "$CANONICAL_UNSAFE_REPO/.kontourai/flow/runs/canonical-unsafe/state.json"
+if FLOW_AGENTS_GOAL_FIT_STRICT=true node "$ROOT/scripts/hooks/stop-goal-fit.js" >"$TMPDIR_EVAL/canonical-malformed.out" 2>"$TMPDIR_EVAL/canonical-malformed.err" <<JSON
+{"hook_event_name":"Stop","cwd":"$CANONICAL_UNSAFE_REPO"}
+JSON
+then
+  _fail "strict goal-fit hook allowed an active scoped session with malformed canonical Flow state"
+elif [[ "$?" -eq 2 ]] && grep -q 'canonical Flow state is unsafe or malformed' "$TMPDIR_EVAL/canonical-malformed.err"; then
+  _pass "strict goal-fit hook fails closed for malformed canonical Flow state of an active scoped session"
+else
+  _fail "strict goal-fit hook did not report malformed canonical Flow state: $(cat "$TMPDIR_EVAL/canonical-malformed.err")"
+fi
+
+if cmp -s "$ROOT/scripts/hooks/stop-goal-fit.js" "$ROOT/context/scripts/hooks/stop-goal-fit.js"; then
+  _pass "canonical Stop hook source and shipped context mirror remain byte-identical"
+else
+  _fail "canonical Stop hook source and shipped context mirror diverged"
+fi
+
 
 if [[ "$errors" -eq 0 ]]; then
   echo "Goal Fit hook integration passed."

package/evals/integration/test_install_merge.sh

@@ -22,6 +22,10 @@ set -euo pipefail
 
 ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)"
 TMPDIR_EVAL="$(mktemp -d /tmp/install-merge.XXXXXX)"
+export HOME="$TMPDIR_EVAL/hermetic-home"
+export FLOW_AGENTS_SKILLS_DIR="$TMPDIR_EVAL/hermetic-universal/skills"
+mkdir -p "$HOME/.agents" "$FLOW_AGENTS_SKILLS_DIR"
+printf 'installer tests must not replace this file\n' > "$HOME/.agents/ambient-sentinel"
 pass=0
 fail=0
 
@@ -1138,7 +1142,8 @@ echo ""
 echo "--- CH4: codex-home preserves user config, profiles, hooks, local kit state, and user skills ---"
 
 CH4_DEST="$TMPDIR_EVAL/codex-home-ch4"
-mkdir -p "$CH4_DEST/kits/local/repositories/user-kit" "$CH4_DEST/sk""ills/user-owned-skill" "$CH4_DEST/ag""ents"
+CH4_SKILLS="$TMPDIR_EVAL/codex-home-ch4-universal-skills"
+mkdir -p "$CH4_DEST/kits/local/repositories/user-kit" "$CH4_SKILLS/user-owned-skill" "$CH4_DEST/ag""ents"
 
 CH4_GENERIC_DIRS=(agent-cards build context docs evals integrations packaging powers prompts schemas scripts kits)
 for CH4_DIR in "${CH4_GENERIC_DIRS[@]}"; do
@@ -1197,14 +1202,14 @@ JSON
 cat > "$CH4_DEST/kits/local/repositories/user-kit/kit.json" << 'JSON'
 {"id":"user-kit","name":"User Kit"}
 JSON
-cat > "$CH4_DEST/sk""ills/user-owned-skill/SKILL.md" << 'EOF_SKILL'
+cat > "$CH4_SKILLS/user-owned-skill/SKILL.md" << 'EOF_SKILL'
 # User Owned Skill
 EOF_SKILL
 cat > "$CH4_DEST/ag""ents/user-owned-agent.toml" << 'EOF_AGENT'
 name = "user-owned-agent"
 EOF_AGENT
 
-CODEX_REAL_HOME="$TMPDIR_EVAL/fake-real-codex" bash "$ROOT_DIR/scripts/install-codex-home.sh" "$CH4_DEST" >/dev/null 2>&1
+CODEX_REAL_HOME="$TMPDIR_EVAL/fake-real-codex" FLOW_AGENTS_SKILLS_DIR="$CH4_SKILLS" bash "$ROOT_DIR/scripts/install-codex-home.sh" "$CH4_DEST" >/dev/null 2>&1
 
 if grep -q 'user-model' "$CH4_DEST/config.toml" \
   && grep -q 'user-profile-model' "$CH4_DEST/custom.config.toml" \
@@ -1243,10 +1248,10 @@ else
   _fail "CH4: kits/local registry or repository was removed"
 fi
 
-if [[ -f "$CH4_DEST/sk""ills/search-first/SKILL.md" && -f "$CH4_DEST/sk""ills/plan-work/SKILL.md" && -f "$CH4_DEST/sk""ills/user-owned-skill/SKILL.md" ]]; then
-  _pass "CH4: root, Builder Kit workflow, and user-owned skills coexist in flattened Codex skills"
+if [[ -f "$CH4_SKILLS/search-first/SKILL.md" && -f "$CH4_SKILLS/plan-work/SKILL.md" && -f "$CH4_SKILLS/user-owned-skill/SKILL.md" ]]; then
+  _pass "CH4: root, Builder Kit workflow, and user-owned skills coexist in universal Codex skills"
 else
-  _fail "CH4: root skill, Builder Kit workflow skill, or user-owned skill missing from flattened Codex skills"
+  _fail "CH4: root skill, Builder Kit workflow skill, or user-owned skill missing from universal Codex skills"
 fi
 
 if [[ -f "$CH4_DEST/scripts/install-merge.js" && -f "$CH4_DEST/build/src/cli/kit.js" && -f "$CH4_DEST/ag""ents/tool-worker.toml" && -f "$CH4_DEST/ag""ents/user-owned-agent.toml" && -f "$CH4_DEST/hooks.json" ]] \
@@ -1272,7 +1277,7 @@ else
   _fail "CH4: installed scripts/kit.js could not run with installed build bundle"
 fi
 
-CODEX_REAL_HOME="$TMPDIR_EVAL/fake-real-codex" bash "$ROOT_DIR/scripts/install-codex-home.sh" "$CH4_DEST" >/dev/null 2>&1
+CODEX_REAL_HOME="$TMPDIR_EVAL/fake-real-codex" FLOW_AGENTS_SKILLS_DIR="$CH4_SKILLS" bash "$ROOT_DIR/scripts/install-codex-home.sh" "$CH4_DEST" >/dev/null 2>&1
 CH4_GENERIC_PRESERVED=1
 for CH4_DIR in "${CH4_GENERIC_DIRS[@]}"; do
   [[ "$(cat "$CH4_DEST/$CH4_DIR/user-owned.txt" 2>/dev/null)" == "user-owned-$CH4_DIR" ]] || CH4_GENERIC_PRESERVED=0
@@ -1288,7 +1293,7 @@ echo ""
 # ─── codex-home: CH5: destination symlink containment ───────────────────────
 echo "--- CH5: codex-home refuses managed destination symlinks ---"
 
-for CH5_REL in scripts kits sk""ills hooks.json; do
+for CH5_REL in scripts kits hooks.json; do
   CH5_DEST="$TMPDIR_EVAL/codex-home-ch5-${CH5_REL//\//-}"
   CH5_OUTSIDE="$TMPDIR_EVAL/codex-home-ch5-outside-${CH5_REL//\//-}"
   mkdir -p "$CH5_DEST" "$CH5_OUTSIDE"
@@ -1680,6 +1685,11 @@ echo ""
 
 echo ""
 echo "==========================="
+if [[ "$(cat "$HOME/.agents/ambient-sentinel" 2>/dev/null)" == "installer tests must not replace this file" ]]; then
+  _pass "dedicated Codex installer calls remain inside hermetic universal roots"
+else
+  _fail "dedicated Codex installer calls mutated the isolated default-home sentinel"
+fi
 total=$((pass + fail))
 echo "Results: ${pass}/${total} passed, ${fail} failed"
 [[ "$fail" -gt 0 ]] && exit 1

package/evals/integration/test_liveness_verdict.sh

@@ -292,23 +292,20 @@ else
 fi
 
 # ─── G. Static skill-text assertions (AC2) ──────────────────────────────────
-echo "--- G. Static pull-work tiebreaker/loser-release skill-text assertions ---"
-
-require_text "$PULL" '### Post-Claim Conflict Re-check' "pull-work documents the Post-Claim Conflict Re-check subsection (AC2)"
-require_text "$PULL" 'When a double-hold is detected, immediately run the deterministic tiebreaker' "pull-work instructs running the deterministic tiebreaker immediately on a detected double-hold (AC2)"
-require_text "$PULL" 'liveness verdict <subjectId> --json' "pull-work references liveness verdict <subjectId> --json by exact command name (AC1, AC2)"
-require_text "$PULL" 'the same pinned .self_actor. and the same .subjectId. already in scope' "pull-work reuses the same pinned self_actor/subjectId already in scope for the verdict call, never re-derived (AC2)"
-require_text "$PULL" 'an exact-timestamp tie breaks by ascending actor-id string comparison \(.reason: \"tie-actor-lexicographic\".\)' "pull-work documents the exact-timestamp tiebreak and its reason value verbatim (AC1, AC2)"
-require_text "$PULL" 'the SAME verdict for the SAME stream state regardless of which actor invokes it' "pull-work states the verdict is deterministic regardless of which actor invokes it (AC1, AC2)"
-require_text "$PULL" 'If .winner\.actor !== self_actor., this session is the loser' "pull-work's loser branch is keyed on winner.actor !== self_actor (AC2)"
-require_text "$PULL" 'immediately run .npm run workflow:sidecar -- liveness release <subjectId> --actor <self_actor>.' "pull-work's loser branch runs liveness release <subjectId> --actor <self_actor> immediately (AC2)"
-require_text "$PULL" 'extend .post_claim_conflict. with .\{verdict_reason, winner_actor, conceded: true\}.' "pull-work's loser branch extends post_claim_conflict with {verdict_reason, winner_actor, conceded: true} (AC2)"
-require_text "$PULL" 'return to .### 3\. Select Work. to reselect within the same .pull-work. pass' "pull-work's loser branch returns to Select Work to reselect within the same pass (AC2)"
-require_text "$PULL" 'excluding the just-released subject' "pull-work's reselect excludes the just-released subject (AC2)"
-require_text "$PULL" 'If .winner\.actor === self_actor., this session wins' "pull-work's winner branch is keyed on winner.actor === self_actor (AC2)"
-require_text "$PULL" 'record the verdict for transparency \(.\{verdict_reason, winner_actor: self_actor, conceded: false\}.\) in .post_claim_conflict. and proceed normally; do not release' "pull-work's winner branch records the verdict for transparency and proceeds without releasing (AC2)"
-require_text "$PULL" 'closes the .detected but advisory-only. gap ADR 0012 §4 names for THIS session.s own double-hold' "pull-work's honesty note states the verdict+release loop closes the detected-but-advisory-only gap for this session's own double-hold (AC2)"
-require_text "$PULL" 'still does not provide true mutual exclusion across the read-then-write race window itself' "pull-work's honesty note distinguishes the new convergence guarantee from the unchanged read-then-write race residual (AC2)"
+echo "--- G. Provider-neutral pull-work conflict contract assertions ---"
+
+require_text "$PULL" 'deterministic conflict arbitration' "pull-work uses provider conflict arbitration"
+require_text "$PULL" 'same pinned caller identity, subject, and observation' "pull-work reuses one conflict snapshot"
+require_text "$PULL" 'Prefer the earlier claim' "pull-work defines deterministic winner precedence"
+require_text "$PULL" 'exact timestamp tie by ascending' "pull-work defines deterministic tie handling"
+require_text "$PULL" 'same winner from the same state' "pull-work requires actor-independent verdicts"
+require_text "$PULL" 'If the caller loses' "pull-work defines loser behavior"
+require_text "$PULL" 'release or concede its claim immediately' "pull-work makes the loser concede"
+require_text "$PULL" 'conceded: true' "pull-work records loser concession"
+require_text "$PULL" 'exclude the contested subject' "pull-work excludes the lost subject before reselection"
+require_text "$PULL" 'If the caller wins' "pull-work defines winner behavior"
+require_text "$PULL" 'conceded: false' "pull-work records winner transparency"
+require_text "$PULL" 'underlying read-then-write race' "pull-work distinguishes convergence from mutual exclusion"
 
 # ─── H. Hostile-actor injection: winner.actor sanitized in --json and text (F5) ─────
 echo "--- H. Hostile-actor verdict sanitization (F5) ---"

package/evals/integration/test_phase_map_and_gate_claim.sh

@@ -8,7 +8,7 @@
 #   4. record-gate-claim at pull-work step produces builder.pull-work.selected claim (status=verified).
 #   5. A TAMPERED bundle (stored verified, evidence fail) at pull-work step BLOCKS (exit 2)
 #      with the tamper warning naming the declared claimType.
-#   6. A CLEAN record-gate-claim bundle is not reported as false completion;
+#   6. A CLEAN public Builder bundle is not reported as false completion;
 #      its still-active canonical Flow run remains blocked from stopping.
 #
 # Deterministic, no model spend, self-cleaning.
@@ -162,10 +162,14 @@ flow_agents_node "workflow-sidecar" init-plan "$CLAIM_ROOT/gate-claim/gate-claim
   --source-request "Test" --summary "Testing" \
   --timestamp "2026-06-26T00:00:00Z" >/dev/null 2>&1
 
+CLAIM_ARTIFACT="$CLAIM_ROOT/gate-claim/gate-claim--pull-work.md"
+printf '# Pull Work\n\nSelected issue #177 for implementation.\n' > "$CLAIM_ARTIFACT"
+
 if flow_agents_node "workflow-sidecar" record-gate-claim "$CLAIM_ROOT/gate-claim" \
   --status pass \
   --summary "Selected issue #177 for implementation." \
   --expectation selected-work \
+  --evidence-ref-json "{\"kind\":\"artifact\",\"file\":\"$CLAIM_ARTIFACT\",\"summary\":\"Declared durable pull-work artifact for selected-work.\"}" \
   --timestamp "2026-06-26T00:01:00Z" >/dev/null 2>&1; then
   _pass "record-gate-claim exits 0 at pull-work step"
 else
@@ -188,9 +192,18 @@ node -e "
     console.error('expected status=verified, got', target.status);
     process.exit(1);
   }
+  const expectedArtifact = '.kontourai/flow-agents/gate-claim/gate-claim--pull-work.md';
+  if (!fs.existsSync('$CLAIM_ARTIFACT')) {
+    console.error('missing declared durable selected-work artifact:', '$CLAIM_ARTIFACT');
+    process.exit(1);
+  }
+  if (!target.metadata?.artifact_refs?.some(ref => ref.kind === 'artifact' && ref.file === expectedArtifact)) {
+    console.error('selected-work claim does not cite declared artifact:', JSON.stringify(target.metadata?.artifact_refs));
+    process.exit(1);
+  }
 " 2>/dev/null \
-  && _pass "bundle contains builder.pull-work.selected with subjectType=work-item, status=verified" \
-  || _fail "bundle missing or incorrect builder.pull-work.selected claim"
+  && _pass "bundle contains builder.pull-work.selected with declared durable artifact evidence" \
+  || _fail "bundle missing selected-work claim or declared durable artifact evidence"
 
 echo ""
 echo "=== 4b. composed publish-learn gate claim emits builder.pr-open.pull-request ==="
@@ -215,10 +228,14 @@ flow_agents_node "workflow-sidecar" advance-state "$COMPOSED_ROOT/composed-gate-
   --flow-definition builder.build \
   --timestamp "2026-06-26T00:00:30Z" >/dev/null 2>&1
 
+COMPOSED_ARTIFACT="$COMPOSED_ROOT/composed-gate-claim/release.json"
+printf '{"schema_version":"1.0","task_slug":"composed-gate-claim","status":"pr_open"}\n' > "$COMPOSED_ARTIFACT"
+
 if flow_agents_node "workflow-sidecar" record-gate-claim "$COMPOSED_ROOT/composed-gate-claim" \
   --status pass \
   --summary "PR opened with verification evidence." \
   --expectation pull-request-opened \
+  --evidence-ref-json "{\"kind\":\"artifact\",\"file\":\"$COMPOSED_ARTIFACT\",\"summary\":\"Declared durable release artifact for pull-request-opened.\"}" \
   --timestamp "2026-06-26T00:01:00Z" >/dev/null 2>&1; then
   _pass "record-gate-claim exits 0 at composed pr-open step"
 else
@@ -244,9 +261,18 @@ node -e "
     console.error('expected status=verified, got', target.status);
     process.exit(1);
   }
+  const expectedArtifact = '.kontourai/flow-agents/composed-gate-claim/release.json';
+  if (!fs.existsSync('$COMPOSED_ARTIFACT')) {
+    console.error('missing declared durable pr-open artifact:', '$COMPOSED_ARTIFACT');
+    process.exit(1);
+  }
+  if (!target.metadata?.artifact_refs?.some(ref => ref.kind === 'artifact' && ref.file === expectedArtifact)) {
+    console.error('pr-open claim does not cite declared artifact:', JSON.stringify(target.metadata?.artifact_refs));
+    process.exit(1);
+  }
 " 2>/dev/null \
-  && _pass "composed bundle contains builder.pr-open.pull-request with active parent flow" \
-  || _fail "composed bundle missing or incorrect builder.pr-open.pull-request claim"
+  && _pass "composed bundle contains builder.pr-open.pull-request with declared durable artifact evidence" \
+  || _fail "composed bundle missing pr-open claim or declared durable artifact evidence"
 
 # ─── Tamper-blocks: stored verified + evidence fail → BLOCK (exit 2) ─────────
 echo ""
@@ -355,59 +381,58 @@ fi
 
 # ─── Clean gate-claim: no false completion, active Flow still blocks ─────────
 echo ""
-echo "=== 6. CLEAN record-gate-claim remains governed by active canonical Flow ==="
+echo "=== 6. CLEAN public Builder evidence remains governed by active canonical Flow ==="
 
 C_DIR="$TMP/clean-test"
+CLEAN_SLUG="provider-clean"
 mkdir -p "$C_DIR"
 printf '# Repo\n' > "$C_DIR/AGENTS.md"
+mkdir -p "$C_DIR/.kontourai/flow-agents/$CLEAN_SLUG"
+CLEAN_ARTIFACT="$C_DIR/.kontourai/flow-agents/$CLEAN_SLUG/$CLEAN_SLUG--pull-work.md"
+printf 'Selected Work Item: provider:clean\n' > "$CLEAN_ARTIFACT"
 
-flow_agents_node "workflow-sidecar" ensure-session \
+if ! FLOW_AGENTS_ACTOR=clean-fixture-actor node "$ROOT/build/src/cli.js" workflow start \
   --artifact-root "$C_DIR/.kontourai/flow-agents" \
-  --task-slug clean \
+  --flow builder.build \
+  --work-item provider:clean \
+  --assignment-provider local-file \
   --title "Clean Test" \
-  --summary "Testing clean gate claim." \
-  --flow-id builder.build \
-  --skip-ownership-guard \
-  --timestamp "2026-06-26T00:00:00Z" >/dev/null 2>&1
+  --summary "Testing clean gate claim." >"$TMP/clean-start.out" 2>&1; then
+  _fail "public clean fixture failed to start: $(cat "$TMP/clean-start.out")"
+fi
 
-flow_agents_node "workflow-sidecar" init-plan "$C_DIR/.kontourai/flow-agents/clean/clean--deliver.md" \
+flow_agents_node "workflow-sidecar" init-plan "$C_DIR/.kontourai/flow-agents/$CLEAN_SLUG/$CLEAN_SLUG--deliver.md" \
   --source-request "Test" --summary "Testing" \
   --timestamp "2026-06-26T00:00:00Z" >/dev/null 2>&1
 
-flow_agents_node "workflow-sidecar" advance-state "$C_DIR/.kontourai/flow-agents/clean" \
-  --status in_progress \
-  --phase pickup \
-  --summary "In progress." \
-  --next-action "done" \
-  --flow-definition builder.build \
-  --timestamp "2026-06-26T00:00:30Z" >/dev/null 2>&1
-
 # Fix next_action so it reads as "done" for the gate
 node -e "
   const fs = require('fs');
-  const f = '$C_DIR/.kontourai/flow-agents/clean/state.json';
+  const f = '$C_DIR/.kontourai/flow-agents/$CLEAN_SLUG/state.json';
   const s = JSON.parse(fs.readFileSync(f, 'utf8'));
   s.next_action = { status: 'done', summary: 'Work complete.' };
   s.status = 'verified';
+  delete s.flow_run;
   fs.writeFileSync(f, JSON.stringify(s, null, 2) + '\n');
 " 2>/dev/null
-
-flow_agents_node "workflow-sidecar" record-gate-claim "$C_DIR/.kontourai/flow-agents/clean" \
-  --status pass \
-  --summary "Selected issue #177 for implementation." \
-  --expectation selected-work \
-  --timestamp "2026-06-26T00:01:00Z" >/dev/null 2>&1
-
-set +e
-clean_out="$(FLOW_AGENTS_GOAL_FIT_MODE=block FLOW_AGENTS_GOAL_FIT_BACKSTOP=skip \
-    node "$GATE" 2>&1 <<< "{\"hook_event_name\":\"Stop\",\"cwd\":\"$C_DIR\"}")"
-clean_exit="$?"
-set -e
-
-if [ "$clean_exit" -eq 2 ] && echo "$clean_out" | grep -q 'release skipped for active Flow run "clean"'; then
-  _pass "clean claim advances to the next Flow step and active run blocks premature Stop"
+rm -f "$C_DIR/.kontourai/flow-agents/$CLEAN_SLUG/$CLEAN_SLUG--deliver.md"
+
+clean_out=""
+clean_exit=0
+for attempt in 1 2; do
+  set +e
+  attempt_out="$(FLOW_AGENTS_GOAL_FIT_MODE=block FLOW_AGENTS_GOAL_FIT_MAX_BLOCKS=2 FLOW_AGENTS_GOAL_FIT_BACKSTOP=skip \
+      node "$GATE" 2>&1 <<< "{\"hook_event_name\":\"Stop\",\"cwd\":\"$C_DIR\"}")"
+  attempt_exit="$?"
+  set -e
+  clean_out="$clean_out$attempt_out"
+  if [ "$attempt_exit" -ne 2 ]; then clean_exit="$attempt_exit"; fi
+done
+
+if [ "$clean_exit" -eq 0 ] && echo "$clean_out" | grep -q 'canonical Flow run remains active at step design-probe'; then
+  _pass "canonical Flow blocks repeated Stop after the sidecar projection and Markdown are removed"
 else
-  _fail "clean active Flow run returned the wrong Stop decision (exit $clean_exit): $clean_out"
+  _fail "canonical Flow could be hidden or auto-released after sidecar rewrite (exit $clean_exit): $clean_out"
 fi
 
 if echo "$clean_out" | grep -q "caught false-completion"; then