feat(execution-playbooks): add plan-execute@1 with a pre-dispatch plan gate

.agents/skills/execution-playbooks/references/plan-execute-v1.md

@@ -0,0 +1,34 @@
+# Execution playbook: `plan-execute@1`
+
+## Accepted brief form
+
+This contract accepts `strike` only, and it applies when the resolved worker lane is an execution-class model: a lower-context lane that carries out a plan instead of designing one.
+The brief must carry a complete `## Execution plan` section, materialized from the planning artifact - the TLC `data/<plan>/tasks/T0X.md` contract when one exists - and `bin/sq-plan-validate.sh` structurally validates it before dispatch.
+The planner owns the design; the executing worker receives the plan, not the sources it was derived from.
+
+## Required sequence
+
+1. Accept a brief whose `## Execution plan` section is complete. The section must carry these five labels, each written as a label line followed by its list entries: `Files to touch` (at least one path-like entry, i.e. real file paths), `Ordered steps` (steps numbered, e.g. `1.`), `Acceptance criteria`, `Verification commands` (a command to run), and `Out of scope` (an explicit out-of-scope statement).
+2. Execute the plan as written; do not redesign, re-scope, or re-plan the change.
+3. Follow `tlc-implement` for the implementation method and write the checklist under `data/<id>/artifacts/`.
+4. Stop with `blocked:` when the plan is incomplete, internally contradictory, or contradicted by the code, naming the exact gap rather than filling it by inference.
+5. Report acceptance evidence and the verification command result; leave review, fixes, delivery, and merge to the existing owners.
+
+## Required evidence
+
+- The brief's materialized `## Execution plan` section, with every required field present and non-empty.
+- A checklist under `data/<id>/artifacts/` recording the plan fields as realized: files touched as planned, steps executed, acceptance met, and the verification command run.
+- A `blocked:` report naming the missing or contradictory plan element when execution cannot proceed safely.
+
+## Exit predicate
+
+Every planned file and step is realized or explicitly reported, acceptance criteria are met, and the planned verification command has been run with its result recorded.
+
+## Stop conditions
+
+Stop when the plan is incomplete or internally contradictory, when a step contradicts the current code, when an unplanned design decision would be required, or when the change would materially exceed the materialized plan.
+
+## Ownership and anti-patterns
+
+The planner (with the commander) owns design, and `AGENTS.md` section 7 plus `execution-playbooks` own selection; this contract only carries the materialized plan into execution.
+Do not redesign the change, silently widen scope, edit or defer the plan instead of executing it, or create playbook-owned review, delivery, approval, or terminal authority.

AGENTS.md

@@ -358,6 +358,7 @@ Once ownership is settled, validate exactly once against that final head so no o
 
 An ask-user finding returns as `needs-decision`; Squad decides only when the configured authority permits, otherwise escalates to the commander.
 For a strike brief with an explicit execution playbook, run its structural validator before sending the implementation to drill or the selected delivery path; the playbook validator never replaces drill.
+When the resolved worker lane is an execution-class model that must not design, select `plan-execute@1` with the plan materialized from the planning artifact and run `bin/sq-plan-validate.sh <id>` before dispatch; `bin/sq-spawn.sh` refuses an incomplete plan.
 Versioned planning and evaluation playbooks are methods only, while delivery remains owned by mode and drill and merge remains commander-authorized.
 `orchestrate` is rejected because Commander, XO, backlog, and supervision already own programme coordination; `autopilot-full`, `autopilot-stack`, and `autonomous-run` remain deferred until their documented measurable triggers are proven, and none adds merge, discard, or destructive authority.
 `arena` means competing approaches to one problem with one selected base, `swarm` means parallel coverage of distinct slices, and `interrogate` means independent attacks on one artifact with deduplication and judgment; these are auxiliary topologies, not playbook identities, dispatch owners, or inferences from free-form `sq-tasks` `kind`.

bin/sq-brief.sh

@@ -187,6 +187,7 @@ if [ "$PLAYBOOK_SET" -eq 1 ]; then
     visual-parity@1) PLAYBOOK_ID=visual-parity; PLAYBOOK_VERSION=1; PLAYBOOK_SECTION=$(cat "$SQUAD_ROOT/.agents/skills/execution-playbooks/references/visual-parity-v1.md"); EXPECTED_KIND='recon|strike' ;;
     multi-phase-plan@1) PLAYBOOK_ID=multi-phase-plan; PLAYBOOK_VERSION=1; PLAYBOOK_SECTION=$(cat "$SQUAD_ROOT/.agents/skills/execution-playbooks/references/multi-phase-plan-v1.md"); EXPECTED_KIND=recon ;;
     "eval@1") PLAYBOOK_ID='eval'; PLAYBOOK_VERSION=1; PLAYBOOK_SECTION=$(cat "$SQUAD_ROOT/.agents/skills/execution-playbooks/references/eval-v1.md"); EXPECTED_KIND=recon ;;
+    plan-execute@1) PLAYBOOK_ID=plan-execute; PLAYBOOK_VERSION=1; PLAYBOOK_SECTION=$(cat "$SQUAD_ROOT/.agents/skills/execution-playbooks/references/plan-execute-v1.md"); EXPECTED_KIND=strike ;;
     orchestrate@1) echo "error: orchestrate@1 disposition=reject; Commander, XO, backlog, and supervision already own programme coordination" >&2; exit 1 ;;
     autopilot-full@1|autopilot-stack@1) echo "error: $PLAYBOOK disposition=defer; re-evaluate only for a real multi-PR programme where commander merge approval is the measured bottleneck and yolo cannot cover it; no auto-merge or drill bypass" >&2; exit 1 ;;
     autonomous-run@1) echo "error: autonomous-run disposition=defer; re-evaluate only with a real task whose cycle the current supervision cannot conduct, plus a verifiable terminal predicate, budget, stop conditions, and duplicate prevention" >&2; exit 1 ;;

bin/sq-plan-validate.sh

@@ -0,0 +1,106 @@
+#!/usr/bin/env bash
+# Deterministic pre-dispatch structural gate for a brief's execution plan.
+# Usage: sq-plan-validate.sh <task-id>
+#
+# Reads data/<id>/brief.md, bounds its `## Execution plan` section, and refuses
+# when a required field is absent or structurally empty. Required fields and
+# shapes (label line, then its entries as a list; short fields may carry inline
+# content after the label):
+#   - Files to touch, with at least one path-like entry
+#   - Ordered steps, with at least one numbered step
+#   - Acceptance criteria
+#   - Verification command(s)
+#   - Out of scope
+# The check is structural only: it never judges plan quality, correctness, or
+# feasibility. It is the dispatch-time companion to bin/sq-playbook-validate.sh,
+# which owns the post-work checklist evidence for the plan-execute@1 playbook.
+set -eu
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+BASE="${SQUAD_BASE:-${SQUAD_HOME:-$ROOT}}"
+DATA="${SQUAD_DATA_OVERRIDE:-$BASE/data}"
+ID=${1:-}
+[ -n "$ID" ] || { echo "error: task id is required" >&2; exit 2; }
+BRIEF="$DATA/$ID/brief.md"
+[ -f "$BRIEF" ] || { echo "missing: $BRIEF"; exit 1; }
+
+awk -v id="$ID" '
+function label_text(l) {
+  sub(/^[[:space:]]+/, "", l)
+  sub(/^[*_]+/, "", l)
+  sub(/[*_]+[[:space:]]*$/, "", l)
+  return l
+}
+function field(l) {
+  if (l ~ /^[Ff]iles[[:space:]]+to[[:space:]]+touch/) return "files"
+  if (l ~ /^([Oo]rdered[[:space:]]+)?[Ss]teps/) return "steps"
+  if (l ~ /^[Aa]cceptance[[:space:]]+criteria/) return "acceptance"
+  if (l ~ /^[Vv]erification[[:space:]]+commands?/) return "verification"
+  if (l ~ /^[Oo]ut[[:space:]]+of[[:space:]]+scope/) return "out-of-scope"
+  return ""
+}
+function flush() {
+  if (cur != "") { seen[cur] = 1; body[cur] = body[cur] "\n" text }
+}
+BEGIN { inplan = 0; heading = 0; cur = ""; text = "" }
+/^##[[:space:]]+[Ee]xecution[[:space:]]+[Pp]lan[[:space:]]*$/ { inplan = 1; heading = 1; next }
+inplan && /^#[[:space:]]/ { inplan = 0; next }
+inplan && /^##[[:space:]]/ { inplan = 0; next }
+!inplan { next }
+{
+  l = label_text($0)
+  f = field(l)
+  if (f != "") {
+    flush()
+    cur = f; text = ""; inline[cur] = ""
+    if (l ~ /:/) {
+      sub(/^[^:]*:/, "", l)
+      gsub(/^[[:space:]]+|[[:space:]]+$/, "", l)
+      inline[cur] = l
+    }
+    next
+  }
+  if (cur != "") {
+    text = text "\n" $0
+    item = $0
+    sub(/^[[:space:]]+/, "", item)
+    if (item ~ /^[-*+][[:space:]]/ || item ~ /^[0-9]+[.)]/) has_item[cur] = 1
+    if (item ~ /^[0-9]+[.)]/) has_step[cur] = 1
+  }
+}
+END {
+  flush()
+  if (!heading) { print "missing: execution plan section (## Execution plan)"; exit 1 }
+  labels["files"] = "files to touch (exact paths)"
+  labels["steps"] = "ordered steps"
+  labels["acceptance"] = "acceptance criteria"
+  labels["verification"] = "verification command"
+  labels["out-of-scope"] = "out of scope"
+  split("files steps acceptance verification out-of-scope", keys, " ")
+  bad = 0
+  for (i = 1; i <= 5; i++) {
+    k = keys[i]
+    if (!(k in seen)) {
+      printf "missing: execution plan field \"%s\"\n", labels[k]
+      bad = 1
+      continue
+    }
+    combined = inline[k] "\n" body[k]
+    if (inline[k] == "" && has_item[k] != 1) {
+      printf "empty: execution plan field \"%s\"\n", labels[k]
+      bad = 1
+      continue
+    }
+    if (k == "files" && combined !~ /\/|[A-Za-z0-9_-]+\.[A-Za-z0-9]+/) {
+      printf "invalid: execution plan field \"%s\" has no path-like entry\n", labels[k]
+      bad = 1
+    }
+    if (k == "steps" && has_step[k] != 1 && inline[k] !~ /^[0-9]+[.)]/) {
+      printf "invalid: execution plan field \"%s\" has no numbered step\n", labels[k]
+      bad = 1
+    }
+  }
+  if (bad) exit 1
+  printf "execution plan structurally valid: %s\n", id
+}
+' "$BRIEF"

bin/sq-playbook-validate.sh

@@ -15,7 +15,7 @@ ARTIFACTS="$DATA/$ID/artifacts"
 PLAYBOOK=$(sed -n 's/^playbook=//p' "$META" | head -n 1)
 VERSION=$(sed -n 's/^playbook_version=//p' "$META" | head -n 1)
 case "$PLAYBOOK@$VERSION" in
-  bug-fix@1|investigation@1|feature@1|refactoring@1|prototype@1|perf@1|hillclimb@1|runtime-forensics@1|trace-forensics@1|visual-parity@1|multi-phase-plan@1|eval@1) ;;
+  bug-fix@1|investigation@1|feature@1|refactoring@1|prototype@1|perf@1|hillclimb@1|runtime-forensics@1|trace-forensics@1|visual-parity@1|multi-phase-plan@1|eval@1|plan-execute@1) ;;
   *) echo "error: task $ID has unsupported execution playbook identity"; exit 1 ;;
 esac
 CHECKLIST=
@@ -172,6 +172,16 @@ case "$PLAYBOOK@$VERSION" in
       "recommendation;promote|reject;production|promotion"
     )
     ;;
+  plan-execute@1)
+    CRITERIA_LABELS=("files touched as planned" "steps executed" "acceptance met" "verification run")
+    CRITERIA_PATTERNS=("files" "steps" "acceptance" "verification")
+    CRITERIA_CHECKS=(
+      "files|paths;planned|materialized"
+      "steps;executed|followed"
+      "acceptance;met|verified|criteria"
+      "verification;run|executed|result"
+    )
+    ;;
 esac
 COUNT=${#CRITERIA_LABELS[@]}
 # Collect all proofs first for duplicate detection.

bin/sq-spawn.sh

@@ -1452,11 +1452,21 @@ if [ "$PLAYBOOK_LINES" -eq 1 ]; then
     'Execution playbook: id=feature version=1') PLAYBOOK_META=feature; PLAYBOOK_VERSION_META=1; EXPECTED_KIND=strike ;;
     'Execution playbook: id=refactoring version=1') PLAYBOOK_META=refactoring; PLAYBOOK_VERSION_META=1; EXPECTED_KIND=strike ;;
     'Execution playbook: id=prototype version=1') PLAYBOOK_META=prototype; PLAYBOOK_VERSION_META=1; EXPECTED_KIND=recon ;;
+    'Execution playbook: id=plan-execute version=1') PLAYBOOK_META=plan-execute; PLAYBOOK_VERSION_META=1; EXPECTED_KIND=strike ;;
     *) echo "error: malformed or unsupported execution playbook identity in $BRIEF" >&2; exit 1 ;;
   esac
   [ "$KIND" = "$EXPECTED_KIND" ] || { echo "error: $PLAYBOOK_META@1 is compatible only with kind: $EXPECTED_KIND" >&2; exit 1; }
   # shellcheck disable=SC2016 # Backticks are literal brief syntax.
   grep -q "^# Execution playbook: \`$PLAYBOOK_META@$PLAYBOOK_VERSION_META\`$" "$BRIEF" || { echo "error: $PLAYBOOK_META@$PLAYBOOK_VERSION_META identity has no materialized contract" >&2; exit 1; }
+  # A plan-execute brief carries the design as a materialized plan: the structural
+  # gate refuses dispatch before any endpoint exists when that plan is incomplete.
+  if [ "$PLAYBOOK_META" = plan-execute ]; then
+    if ! plan_validate_out=$(SQUAD_BASE="$SQUAD_BASE" "$SCRIPT_DIR/sq-plan-validate.sh" "$ID" 2>&1); then
+      printf '%s\n' "$plan_validate_out" >&2
+      echo "error: $ID brief execution plan failed structural validation; materialize a complete plan before dispatch" >&2
+      exit 1
+    fi
+  fi
 fi
 
 delivery_rigor_rank() {  # <mode> -> 3 (most rigor) .. 1 (least); 0 = not a task mode

docs/documentation-audiences.json

@@ -1564,6 +1564,10 @@
    "path": ".agents/skills/execution-playbooks/references/eval-v1.md",
    "audience": "agent-runtime"
   },
+  {
+   "path": ".agents/skills/execution-playbooks/references/plan-execute-v1.md",
+   "audience": "agent-runtime"
+  },
   {
    "path": "skills/interview-me/SKILL.md",
    "audience": "public-product"

docs/pt-BR/scripts.md

@@ -127,7 +127,7 @@ A recusa compartilhada do gate drill para entrypoints do ciclo de vida da unidad
 | `sq-pr-check-migrate.sh` | Quarentenair polls de tarefa antigos sem execução e reconstruir apenas polls canônicos       |
 | `sq-pr-check.sh`         | Registrar valores validados de `pr=` e `pr_head=`, então armar atomicamente um poll estático de merge |
 | `sq-pr-merge.sh`         | Registrar metadados do PR, então mesclar a URL completa canônica GitHub do PR da tarefa                    |
-| `sq-playbook-validate.sh` | Validar evidência estrutural de um playbook de execução materializado (bug-fix@1, investigation@1, feature@1, refactoring@1, prototype@1, perf@1, hillclimb@1, runtime-forensics@1, trace-forensics@1, visual-parity@1, multi-phase-plan@1, eval@1) |
+| `sq-playbook-validate.sh` | Validar evidência estrutural de um playbook de execução materializado (bug-fix@1, investigation@1, feature@1, refactoring@1, prototype@1, perf@1, hillclimb@1, runtime-forensics@1, trace-forensics@1, visual-parity@1, multi-phase-plan@1, eval@1, plan-execute@1) |
 | `sq-promote.sh`          | Promover uma tarefa recon in-place a uma tarefa strike protegida com modo de entrega explícito |
 | `sq-teardown.sh`         | Teardown fail-closed: devolver worktrees ship landadas, exigir entregáveis completos de recon, aposentar bases XO |
 | `sq-harness.sh`          | Detectar o harness em execução e resolver crew ou XO harness, modelo e esforço |

docs/verification/playbook-absorptions.md

@@ -20,7 +20,7 @@ On 2026-09-19, the focused `sq-brief` test exercised `--playbook <name>@1` for e
 sq-brief.sh: absorbed playbook names are refused and leave no selectable brief
 ```
 
-The test is [`test_absorbed_playbook_names_are_not_selectable`](../../tests/sq-brief.test.sh), and the registry contains wave-1 contracts `bug-fix@1`, `feature@1`, `investigation@1`, `prototype@1`, `refactoring@1` plus wave-2 contracts `perf@1`, `hillclimb@1`, `runtime-forensics@1`, `trace-forensics@1`, and `visual-parity@1` plus planning/evaluation methods `multi-phase-plan@1` and `eval@1` under `.agents/skills/execution-playbooks/references/`.
+The test is [`test_absorbed_playbook_names_are_not_selectable`](../../tests/sq-brief.test.sh), and the registry contains wave-1 contracts `bug-fix@1`, `feature@1`, `investigation@1`, `prototype@1`, `refactoring@1` plus wave-2 contracts `perf@1`, `hillclimb@1`, `runtime-forensics@1`, `trace-forensics@1`, and `visual-parity@1` plus planning/evaluation methods `multi-phase-plan@1` and `eval@1` plus the materialized-plan execution contract `plan-execute@1` under `.agents/skills/execution-playbooks/references/`.
 
 The focused test command is:
 
@@ -34,7 +34,7 @@ The registry inventory command is:
 find .agents/skills/execution-playbooks -maxdepth 2 -type f -print | sort
 ```
 
-Its expected output contains `SKILL.md` and the reference files for all twelve selectable execution playbook contracts.
+Its expected output contains `SKILL.md` and the reference files for all thirteen selectable execution playbook contracts.
 
 ## Deferred and auxiliary policy
 
@@ -49,4 +49,4 @@ The three auxiliary topologies are not execution playbook identities or dispatch
 `arena` competes on one problem and selects one base, `swarm` covers distinct slices, and `interrogate` independently attacks one artifact with deduplication and judgment.
 For PR or diff surfaces, `interrogate` points to the read-only Drill surface documented in [`docs/pr-review.md`](../pr-review.md), and separate reviews are limited to requested or knowledge-only review deliverables.
 
-The catalog remains 22 playbooks because these topology names and the four refused or deferred names are not selectable identities.
+The catalog now counts 23 playbook identities: 13 selectable contracts, 5 absorbed upstream names, and 5 refused or deferred names.

tests/sq-brief.test.sh

@@ -405,14 +405,27 @@ test_remaining_lifecycle_playbooks() {
   assert_grep "not a layer-by-layer" "$brief" "multi-phase-plan contract missing layer boundary"
   assert_grep "existing backlog" "$brief" "multi-phase-plan contract missing backlog handoff"
 
+  SQUAD_BASE="$home" "$ROOT/bin/sq-brief.sh" lifecycle-plan-execute repo --mode drill --playbook plan-execute@1 >/dev/null 2>&1 || fail "plan-execute@1 should materialize as strike"
+  brief="$home/data/lifecycle-plan-execute/brief.md"
+  assert_grep "Execution playbook: id=plan-execute version=1" "$brief" "plan-execute identity missing"
+  assert_grep "execution-class" "$brief" "plan-execute contract missing execution-class scope"
+  assert_grep "materialized from the planning artifact" "$brief" "plan-execute contract missing the plan materialization rule"
+  assert_grep "tlc-implement" "$brief" "plan-execute contract missing tlc-implement"
+  assert_grep "data/<id>/artifacts/" "$brief" "plan-execute contract missing the checklist location"
+  assert_grep "blocked:" "$brief" "plan-execute contract missing the stop-with-blocked rule"
+  out=$(SQUAD_BASE="$home" "$ROOT/bin/sq-brief.sh" lifecycle-plan-execute-recon repo --recon --playbook plan-execute@1 2>&1); status=$?
+  [ "$status" -ne 0 ] || fail "plan-execute@1 must refuse recon"
+  assert_contains "$out" "accepts only kind: strike" "plan-execute refusal should name strike compatibility"
+  assert_absent "$home/data/lifecycle-plan-execute-recon/brief.md" "plan-execute recon refusal must not leave a partial brief"
+
   SQUAD_BASE="$home" "$ROOT/bin/sq-brief.sh" lifecycle-eval repo --recon --playbook eval@1 >/dev/null 2>&1 || fail "eval@1 should materialize"
   brief="$home/data/lifecycle-eval/brief.md"
   assert_grep "Execution playbook: id=eval version=1" "$brief" "eval identity missing"
   assert_grep "candidate-visible" "$brief" "eval contract missing candidate blinding"
   assert_grep "chain-elicitation" "$brief" "eval contract missing chain-elicitation prevention"
   assert_grep "does not enter production" "$brief" "eval contract missing promotion boundary"
 
-  for playbook in shipping@2 multi-phase-plan@2 eval@2; do
+  for playbook in shipping@2 multi-phase-plan@2 eval@2 plan-execute@2; do
     out=$(SQUAD_BASE="$home" "$ROOT/bin/sq-brief.sh" "lifecycle-invalid-$RANDOM" repo --mode drill --playbook "$playbook" 2>&1); status=$?
     [ "$status" -ne 0 ] || fail "$playbook should be refused"
     assert_contains "$out" "unknown execution playbook" "$playbook should identify invalid identity"