prizmkit 1.1.101 → 1.1.102

package/bundled/VERSION.json

@@ -1,5 +1,5 @@
 {
-  "frameworkVersion": "1.1.101",
-  "bundledAt": "2026-07-06T10:58:41.729Z",
-  "bundledFrom": "fa258ae"
+  "frameworkVersion": "1.1.102",
+  "bundledAt": "2026-07-06T15:09:00.037Z",
+  "bundledFrom": "e918269"
 }

package/bundled/dev-pipeline/prizmkit_runtime/runner_models.py

@@ -45,6 +45,7 @@ class RunnerEnvironment:
     heartbeat_stale_threshold_seconds: int = 600
     stale_kill_threshold_seconds: int = 600
     max_log_size: int | None = None
+    max_retries: int = 3
     strict_behavior_check: bool = True
 
     @classmethod
@@ -65,6 +66,7 @@ class RunnerEnvironment:
             ),
             stale_kill_threshold_seconds=_int_value(values, "STALE_KILL_THRESHOLD", 600),
             max_log_size=_optional_int(values.get("MAX_LOG_SIZE")),
+            max_retries=max(0, _int_value(values, "MAX_RETRIES", 3)),
             strict_behavior_check=not _falsey(values.get("STRICT_BEHAVIOR_CHECK")),
         )
 
@@ -77,6 +79,7 @@ class RunnerInvocation:
     list_path: Path
     item_id: str = ""
     item_filter: str = ""
+    max_retries: int | None = None
     max_infra_retries: int = 3
     mode: str | None = None
     critic: bool | None = None
@@ -228,6 +231,7 @@ def parse_invocation(
     list_path = family.plan_path
     item_id = ""
     item_filter = ""
+    max_retries: int | None = None
     max_infra_retries = 3
     mode: str | None = None
     critic: bool | None = None
@@ -259,6 +263,11 @@ def parse_invocation(
             item_filter = args[index]
         elif arg.startswith("--features="):
             item_filter = arg.split("=", 1)[1]
+        elif arg == "--max-retries" and index + 1 < len(args):
+            index += 1
+            max_retries = max(0, _safe_int(args[index], 3))
+        elif arg.startswith("--max-retries="):
+            max_retries = max(0, _safe_int(arg.split("=", 1)[1], 3))
         elif arg == "--max-infra-retries" and index + 1 < len(args):
             index += 1
             max_infra_retries = _safe_int(args[index], 3)
@@ -293,6 +302,7 @@ def parse_invocation(
         list_path=list_path,
         item_id=item_id,
         item_filter=item_filter,
+        max_retries=max_retries,
         max_infra_retries=max_infra_retries,
         mode=mode,
         critic=critic,

package/bundled/dev-pipeline/prizmkit_runtime/runner_status.py

@@ -8,7 +8,7 @@ import sys
 from dataclasses import dataclass
 from pathlib import Path
 
-from .runner_models import RunnerFamily, RunnerInvocation
+from .runner_models import RunnerEnvironment, RunnerFamily, RunnerInvocation
 
 
 @dataclass(frozen=True)
@@ -30,6 +30,12 @@ class StatusCommandResult:
         return self.stdout.strip() or self.stderr.strip()
 
 
+def _resolved_max_retries(invocation: RunnerInvocation) -> int:
+    if invocation.max_retries is not None:
+        return max(0, invocation.max_retries)
+    return RunnerEnvironment.from_env().max_retries
+
+
 def run_status_action(
     family: RunnerFamily,
     action: str,
@@ -38,7 +44,7 @@ def run_status_action(
     item_id: str = "",
     session_status: str = "",
     session_id: str = "",
-    max_retries: int = 3,
+    max_retries: int | None = None,
     active_dev_branch: str = "",
     base_branch: str = "",
     last_fatal_error_code: str = "",
@@ -47,7 +53,7 @@ def run_status_action(
     progress_fingerprint: object | None = None,
     project_root: Path | None = None,
 ) -> StatusCommandResult:
-    """Invoke an existing updater script and parse JSON output when present."""
+    max_retries = _resolved_max_retries(invocation) if max_retries is None else max(0, max_retries)
     command = [
         sys.executable,
         str(family.updater_script),

package/bundled/dev-pipeline/prizmkit_runtime/runners.py

@@ -114,6 +114,8 @@ def _run_pipeline(family: RunnerFamily, invocation: RunnerInvocation, paths) ->
         details.append(f"{item_id}:{final_status}")
         if final_status == "completed":
             continue
+        if final_status == "pending":
+            continue
         if RunnerEnvironment.from_env().stop_on_failure:
             return PipelineRunResult(False, processed, "stop_on_failure", final_status, tuple(details))
         if final_status in {"failed", "needs_info"}:
@@ -401,6 +403,7 @@ def _resolve_invocation_paths(invocation: RunnerInvocation, project_root: Path)
         list_path=(project_root / invocation.list_path).resolve(),
         item_id=invocation.item_id,
         item_filter=invocation.item_filter,
+        max_retries=invocation.max_retries,
         max_infra_retries=invocation.max_infra_retries,
         mode=invocation.mode,
         critic=invocation.critic,
@@ -418,5 +421,5 @@ def _help_text(family: RunnerFamily) -> str:
     return (
         f"Python foreground runner for {family.kind}.\n"
         "Supported legacy options include plan list path, item id, --features for feature, "
-        "--max-infra-retries, --mode, --critic/--no-critic, --resume-phase, and --dry-run."
+        "--max-retries, --max-infra-retries, --mode, --critic/--no-critic, --resume-phase, and --dry-run."
     )

package/bundled/dev-pipeline/prizmkit_runtime/sessions.py

@@ -152,7 +152,14 @@ def build_ai_session_command(config: AISessionConfig) -> AISessionCommand:
     platform = _normalize_platform(config.platform)
     prompt = _prompt_text(config.prompt_path)
     if platform == "claude":
-        argv = [config.cli_command, "-p", prompt, "--dangerously-skip-permissions"]
+        argv = [
+            config.cli_command,
+            "-p",
+            prompt,
+            "--dangerously-skip-permissions",
+            "--disallowedTools",
+            "EnterWorktree",
+        ]
         if config.verbose or config.use_stream_json:
             argv.append("--verbose")
         if config.use_stream_json:
@@ -175,6 +182,7 @@ def build_ai_session_command(config: AISessionConfig) -> AISessionCommand:
         timeout = _codex_subagent_timeout(config)
         if timeout > 0:
             argv.extend(["--config", f"agents.job_max_runtime_seconds={timeout}"])
+        argv.extend(["--profile", "no-worktree"])
         argv.extend(["exec", "--cd", str(config.cwd), "--skip-git-repo-check"])
         if config.use_stream_json:
             argv.append("--json")
@@ -186,7 +194,7 @@ def build_ai_session_command(config: AISessionConfig) -> AISessionCommand:
         argv.append("-")
         return AISessionCommand(tuple(argv), prompt, "codex", "stdin")
 
-    argv = [config.cli_command, "--print", "-y"]
+    argv = [config.cli_command, "--print", "-y", "--disallowedTools", "EnterWorktree"]
     if config.verbose:
         argv.append("--verbose")
     if config.use_stream_json:

package/bundled/dev-pipeline/scripts/generate-bootstrap-prompt.py

@@ -676,7 +676,7 @@ def determine_pipeline_mode(complexity):
 SECTION_TO_SKILL = {
     "phase0-init": ("prizmkit-init", "Project Bootstrap",
                     [".prizmkit/prizm-docs/root.prizm", ".prizmkit/config.json"]),
-    "phase0-test-baseline": ("test-baseline", "Test Baseline", []),
+    "phase0-test-baseline": ("test-baseline", "Detect Test Commands", []),
     "phase-context-snapshot": ("context-snapshot", "Build Context Snapshot",
                                [".prizmkit/specs/{slug}/context-snapshot.md"]),
     "phase-specify-plan": ("context-snapshot-and-plan", "Specify & Plan",
@@ -1559,6 +1559,18 @@ def assemble_sections(pipeline_mode, sections_dir, init_done, is_resume,
                           load_section(sections_dir,
                                        "phase-implement-agent.md")))
 
+    # Verification Gates are included in Task Contract. Keep AC in one place so
+    # background context and implementation prompts cannot redefine scope.
+    # --- Review ---
+    if pipeline_mode == "full":
+        sections.append(("phase-review",
+                          load_section(sections_dir,
+                                       "phase-review-full.md")))
+    else:
+        sections.append(("phase-review",
+                          load_section(sections_dir,
+                                       "phase-review-agent.md")))
+
     # --- Test Failure Recovery Protocol (tier-specific) ---
     if pipeline_mode == "lite":
         sections.append(("test-failure-recovery",
@@ -1574,18 +1586,6 @@ def assemble_sections(pipeline_mode, sections_dir, init_done, is_resume,
                       load_section(sections_dir,
                                    "phase-prizmkit-test.md")))
 
-    # Verification Gates are included in Task Contract. Keep AC in one place so
-    # background context and implementation prompts cannot redefine scope.
-    # --- Review (only for agent tiers) ---
-    if pipeline_mode == "full":
-        sections.append(("phase-review",
-                          load_section(sections_dir,
-                                       "phase-review-full.md")))
-    elif pipeline_mode == "standard":
-        sections.append(("phase-review",
-                          load_section(sections_dir,
-                                       "phase-review-agent.md")))
-
     # --- Browser Verification (conditional, tool-aware) ---
     if browser_enabled:
         if browser_tool == "opencli":

package/bundled/dev-pipeline/templates/agent-prompts/dev-implement.md

@@ -61,7 +61,7 @@ Before returning, append `## Implementation Log` to `context-snapshot.md` with:
 - Do not run git commands; staging and commit are handled by the orchestrator.
 - **Edit safety**: If an Edit fails with 'String to replace not found', grep for the target text before retrying. Never guess file offsets — verify them with a Read or grep first.
 - **Read safety**: If 3 consecutive Reads to the same file return 'shorter than offset' or 'Wasted call', STOP and report BLOCKED.
-- **Test early**: Run `{{TEST_CMD}}` after every 3 successful Edit operations. Capture output to /tmp/test-out.txt and grep for failures.
+- **Test gate placement**: Do not run mandatory periodic or full-suite tests here. The scoped/full test gate runs after code review; only use tiny targeted checks when needed for local debugging.
 
 Do not return success unless:
 1. implementation tasks are complete;

package/bundled/dev-pipeline/templates/agent-prompts/reviewer-review.md

@@ -2,6 +2,6 @@
 1. Read `.prizmkit/specs/{{FEATURE_SLUG}}/spec.md` (if it exists) for goals and Verification Gates; if spec.md does not exist, read `.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md` Section 1 Task Contract instead
 2. Read `.prizmkit/specs/{{FEATURE_SLUG}}/plan.md` for architecture decisions and completed tasks
 3. Run /prizmkit-code-review with artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/. The skill will run its internal review-fix loop (Reviewer → filter → Dev fix, max 3 rounds) and write review-report.md.
-4. Run the full test suite using `{{TEST_CMD}}`. When running tests: `({{TEST_CMD}}) 2>&1 | tee /tmp/review-test-out.txt | tail -20`, then grep `/tmp/review-test-out.txt` for details — do NOT re-run the suite multiple times.
+4. Do not run test suites here. The orchestrator runs `/prizmkit-test` after review completes and then fixes code/tests until the scoped gate passes.
 5. review-report.md will be written to .prizmkit/specs/{{FEATURE_SLUG}}/ by prizmkit-code-review.
 Report: verdict (PASS/NEEDS_FIXES), number of rounds, findings fixed/rejected."

package/bundled/dev-pipeline/templates/bootstrap-prompt.md

@@ -30,9 +30,9 @@ Infer what needs to be done from the feature context above and follow the standa
 
 3. **Implement**: Run `/prizmkit-implement` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` to execute the plan using TDD (write tests first, then implement).
 
-4. **Test**: Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` to generate and verify tests only for this feature's changed scope. Do not use bugfix/refactor artifact directories for this gate.
+4. **Review**: Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` to review and auto-fix changes against the spec (internal review-fix loop, max 3 rounds).
 
-5. **Review**: Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` to review and auto-fix changes against the spec (internal review-fix loop, max 3 rounds).
+5. **Test**: Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` to generate and verify tests only for this feature's changed scope after review. Do not use bugfix/refactor artifact directories for this gate.
 
 6. **Retrospective**: Run `/prizmkit-retrospective` to sync `.prizmkit/prizm-docs/` with code changes.
 

package/bundled/dev-pipeline/templates/bootstrap-tier1.md

@@ -145,9 +145,9 @@ You know this project's tech stack. Identify ALL test commands that apply (e.g.,
 **3b.** Run `/prizmkit-implement` — this handles the full implementation cycle:
 - Reads plan.md Tasks section from `.prizmkit/specs/{{FEATURE_SLUG}}/`
 - Reads context from `context-snapshot.md` (Prizm docs, TRAPS, file manifest)
-- Implements task-by-task with TDD, marking each `[x]` immediately
+- Implements task-by-task, marking each `[x]` immediately
 - Creates/updates L2 `.prizm` docs when creating new modules or significantly modifying existing ones — AI selectively decides which modules warrant L2 based on complexity and importance
-- Runs tests using `TEST_CMD` after each task
+- Defers scoped/full test execution until after review
 - Writes '## Implementation Log' to `context-snapshot.md`
 
 **3b-checkpoint.** Context management — if plan.md has more than 5 tasks, update durable checkpoints/artifacts after every 3 tasks, minimize output, and continue. If context overflow occurs in a headless run, the runner starts a fresh continuation session from those checkpoints and artifacts. In an interactive Claude Code session operated manually outside the headless pipeline, `/compact` may be used as an optional convenience only.
@@ -155,10 +155,9 @@ You know this project's tech stack. Identify ALL test commands that apply (e.g.,
 **3c.** After implement completes, verify:
 1. All tasks in plan.md are `[x]`
 2. **Verification Gate Self-Check**: for each acceptance criterion, write one line of evidence (test name / file:line / fixture path). Any gate without evidence is BLOCKED, not success.
-3. **Layered test strategy**: debug with single-file tests (`npx vitest run <file> -t '<name>'`), only run full suite as gate when single-file is green. Never run full suite while a single-file test is still red.
-4. Run the full test suite to ensure nothing is broken
-5. Verify each acceptance criterion from Section 1 of context-snapshot.md is met — check mentally, do NOT re-read files you already wrote
-6. If any criterion is not met, fix it now using the convergence-based recovery loop below
+3. **Layered test strategy**: do not run mandatory periodic/full-suite tests during implementation. Use only tiny targeted checks when needed for local debugging; the scoped feature test gate runs after review.
+4. Verify each acceptance criterion from Section 1 of context-snapshot.md is met — check mentally, do NOT re-read files you already wrote
+5. If any criterion is not met, fix it now using the convergence-based recovery loop below
 
 **CP-2**: All acceptance criteria met, all tests pass.
 
@@ -177,7 +176,7 @@ When tests fail, use convergence recovery — keep fixing while progress is bein
 
 ### Scoped Feature Test Gate — PrizmKit Test
 
-Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` after implementation and before review/commit.
+Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` after review and before commit.
 
 Before invoking the skill, create `.prizmkit/specs/{{FEATURE_SLUG}}/.prizmkit-test-started` so stale reports are ignored.
 

package/bundled/dev-pipeline/templates/bootstrap-tier2.md

@@ -302,7 +302,7 @@ If GATE:MISSING — send message to Dev (re-spawn if needed): "Write the '## Imp
 
 ### Scoped Feature Test Gate — PrizmKit Test
 
-Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` after implementation and before review/commit.
+Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` after review and before commit.
 
 Before invoking the skill, create `.prizmkit/specs/{{FEATURE_SLUG}}/.prizmkit-test-started` so stale reports are ignored.
 

package/bundled/dev-pipeline/templates/bootstrap-tier3.md

@@ -320,7 +320,7 @@ All tasks `[x]`, tests pass.
 
 ### Scoped Feature Test Gate — PrizmKit Test
 
-Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` after implementation and before review/commit.
+Run `/prizmkit-test scope=this-change artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` after review and before commit.
 
 Before invoking the skill, create `.prizmkit/specs/{{FEATURE_SLUG}}/.prizmkit-test-started` so stale reports are ignored.
 

package/bundled/dev-pipeline/templates/sections/phase-implement-agent.md

@@ -19,15 +19,15 @@ grep -c '^\- \[ \]' .prizmkit/specs/{{FEATURE_SLUG}}/plan.md 2>/dev/null || true
 **3b.** Run `/prizmkit-implement` directly — you (the orchestrator) execute the full cycle:
 - Read plan.md Tasks from `.prizmkit/specs/{{FEATURE_SLUG}}/`
 - Use context-snapshot.md Section 4 File Manifest for targeted reads — do NOT re-read files already summarized there
-- Implement task-by-task with TDD, marking each `[x]` immediately
+- Implements task-by-task, marking each `[x]` immediately
 - Creates/updates L2 `.prizm` docs when creating new modules
-- Runs tests using `TEST_CMD`
+- Defers scoped/full test execution until after the code-review loop completes
 - If plan.md has >5 tasks: update checkpoints after every 3 tasks
 
-**3c. Layered test strategy** (replaces "run full suite every 3 Edits"):
-- Debugging a single test: `npx vitest run <test-file> -t '<test name>'` (or per-stack equivalent). Capture to `/tmp/test-out.txt`, grep for failures.
-- Only when single-file is green, run ONE full `{{TEST_CMD}}` as the gate.
-- Never run full suite while a single-file test is still red.
+**3c. Focused checks only (no test gate here)**:
+- During implementation, do not run mandatory periodic or full-suite tests.
+- If a tiny targeted check is needed to understand a local failure, run only the smallest relevant command and capture output to `/tmp/test-out.txt`.
+- The scoped feature test gate runs after code review; do not declare implementation success based on an early full-suite run.
 
 **3d. Verification Gate Self-Check (BLOCKING — before declaring implement done)**:
 For each Verification Gate from the Task Contract, write one evidence line to Implementation Log:
@@ -45,7 +45,7 @@ For each Verification Gate from the Task Contract, write one evidence line to Im
 - Gate Evidence section
 - unresolved blockers
 
-**CP-2**: All tasks `[x]`, Implementation Log written with Gate Evidence, Test Failure Recovery Protocol's Success Rule satisfied. Blocked gates → `failure-log.md`, not success.
+**CP-2**: All tasks `[x]`, Implementation Log written with Gate Evidence, and blocked gates documented in `failure-log.md`. Test execution is deferred until after code review.
 
 **Checkpoint update**:
 ```bash

package/bundled/dev-pipeline/templates/sections/phase-implement-full.md

@@ -20,15 +20,15 @@ Save pre-existing failing tests as `BASELINE_FAILURES`.
 **3b.** Run `/prizmkit-implement` directly — you (the orchestrator) execute the full cycle:
 - Read plan.md Tasks from `.prizmkit/specs/{{FEATURE_SLUG}}/`
 - Use context-snapshot.md Section 4 File Manifest for targeted reads — do NOT re-read files already summarized there
-- Implement task-by-task with TDD, marking each `[x]` immediately
+- Implements task-by-task, marking each `[x]` immediately
 - Creates/updates L2 `.prizm` docs when creating new modules
-- Runs tests using `TEST_CMD`
+- Defers scoped/full test execution until after the code-review loop completes
 - If plan.md has >5 tasks: update checkpoints after every 3 tasks
 
-**3c. Layered test strategy** (replaces "run full suite every 3 Edits"):
-- Debugging a single test: `npx vitest run <test-file> -t '<test name>'` (or per-stack equivalent). Capture to `/tmp/test-out.txt`, grep for failures.
-- Only when single-file is green, run ONE full `{{TEST_CMD}}` as the gate.
-- Never run full suite while a single-file test is still red.
+**3c. Focused checks only (no test gate here)**:
+- During implementation, do not run mandatory periodic or full-suite tests.
+- If a tiny targeted check is needed to understand a local failure, run only the smallest relevant command and capture output to `/tmp/test-out.txt`.
+- The scoped feature test gate runs after code review; do not declare implementation success based on an early full-suite run.
 
 **3d. Verification Gate Self-Check (BLOCKING — before declaring implement done)**:
 For each Verification Gate from the Task Contract, write one evidence line to Implementation Log:
@@ -46,7 +46,7 @@ For each Verification Gate from the Task Contract, write one evidence line to Im
 - Gate Evidence section
 - unresolved blockers
 
-**CP-2**: All tasks `[x]`, Implementation Log written with Gate Evidence, Test Failure Recovery Protocol's Success Rule satisfied. Blocked gates → `failure-log.md`, not success.
+**CP-2**: All tasks `[x]`, Implementation Log written with Gate Evidence, and blocked gates documented in `failure-log.md`. Test execution is deferred until after code review.
 
 **Checkpoint update**:
 ```bash

package/bundled/dev-pipeline/templates/sections/phase-implement-lite.md

@@ -1,33 +1,24 @@
-### Implement + Test
+### Implement
 
 **Protocol references**:
 - Follow Context Budget Rules §8 for scaffold/generated files.
 - Follow Context Budget Rules §9 before package install/build commands that resolve dependencies.
 - Follow Context Budget Rules §10 after build/compile commands.
 
-**3a.** Detect test commands and record baseline:
-
-You know this project's tech stack. Identify ALL test commands that apply (e.g., `go test ./...`, `npm test`, `cargo test`, `pytest`, `make test`, etc.). Record them as `TEST_CMDS`. Then record baseline:
-```bash
-# Run each test command, capture output
-($TEST_CMD) 2>&1 | tee /tmp/test-baseline.txt | tail -20
-```
-
-**3b.** Run `/prizmkit-implement` — this handles the full implementation cycle:
+**3a.** Run `/prizmkit-implement` — this handles the implementation cycle:
 - Reads plan.md Tasks section from `.prizmkit/specs/{{FEATURE_SLUG}}/`
 - Reads context from `context-snapshot.md` (Prizm docs, TRAPS, file manifest)
-- Implements task-by-task with TDD, marking each `[x]` immediately
+- Implements task-by-task, marking each `[x]` immediately
 - Creates/updates L2 `.prizm` docs when creating new modules or significantly modifying existing ones — AI selectively decides which modules warrant L2 based on complexity and importance
-- Runs tests using `TEST_CMD` after each task
+- Does not run the scoped/full test gate; testing happens after code review in the PrizmKit Test phase
 - Writes '## Implementation Log' to `context-snapshot.md`
 
-**3c.** After implement completes, verify:
+**3b.** After implement completes, verify:
 1. All tasks in plan.md are `[x]`
-2. Run the full test suite to ensure no new regressions remain
-3. Verify each Verification Gate from the Task Contract — check mentally, do NOT re-read files you already wrote
-4. If any gate is unmet or blocked, follow the Test Failure Recovery Protocol
+2. Verify each Verification Gate from the Task Contract using code/file evidence already in context — do NOT re-read files you already wrote
+3. If any gate is unmet or blocked, write `failure-log.md` and stop for recovery
 
-**CP-2**: Implementation may proceed only when all tasks are `[x]` and the Test Failure Recovery Protocol's Success Rule is satisfied. Blocked gates must be documented in `failure-log.md` and are not success.
+**CP-2**: Implementation may proceed only when all tasks are `[x]` and blocked gates are documented in `failure-log.md`. Test execution is deferred until after code review.
 
 
 **Checkpoint update**: Run the update script to set step `prizmkit-implement` to `"completed"`:

package/bundled/dev-pipeline/templates/sections/phase-prizmkit-test.md

@@ -1,6 +1,6 @@
 ### Scoped Feature Test Gate — PrizmKit Test
 
-**Goal**: Generate and verify tests for this feature's changed scope before code review without embedding the full gate implementation in this prompt.
+**Goal**: After code review completes, generate and verify tests for this feature's changed scope without embedding the full gate implementation in this prompt.
 
 Create a start marker immediately before invoking the skill so the gate can reject stale reports from older sessions:
 
@@ -34,6 +34,6 @@ python3 "$PIPELINE_DIR/scripts/prizmkit-test-gate.py" check \
 
 Gate outcomes:
 - `GATE:PASS` → append the report path and a 3-5 bullet summary to `context-snapshot.md` under `## PrizmKit Test Gate`, then proceed to the next checkpoint step. The gate script already wrote `test-report-path.txt` and marked `prizmkit-test` completed.
-- `GATE:FAIL` with `NEEDS_FIXES` → read the report's `In-Scope Failures`, fix the feature implementation or generated tests within this feature scope, rerun `/prizmkit-test`, and repeat the gate check. Do not proceed to code review while the verdict is not `PASS`.
+- `GATE:FAIL` with `NEEDS_FIXES` → read the report's `In-Scope Failures`, fix the feature implementation or generated tests within this feature scope, rerun `/prizmkit-test`, and repeat the gate check. Do not proceed to retrospective/commit while the verdict is not `PASS`.
 - `GATE:FAIL` with `BLOCKED`, wrong mode, wrong artifact dir, unreadable report, stale scope, boundary failure, validator failure, or checkpoint update failure → write `.prizmkit/specs/{{FEATURE_SLUG}}/failure-log.md` with the scoped test failure and stop for recovery.
 - `GATE:MISSING` → perform one bounded status check: inspect `/prizmkit-test` output and `.prizmkit/test/` for a report directory. If no current-run report exists, write `failure-log.md` and stop for recovery.

package/bundled/dev-pipeline/templates/sections/phase-review-agent.md

@@ -1,8 +1,6 @@
 ### Review — Code Review
 
-Precondition: the `prizmkit-test` checkpoint is completed and `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt` points to the scoped test report.
-
-Before invoking review, read `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt`, read the referenced `/prizmkit-test` report, and confirm the summary in `context-snapshot.md` under `## PrizmKit Test Gate` matches it. The code-review skill must receive this context through `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` and the Reviewer must assess generated/updated test quality, `In-Scope Failures`, `Baseline Failures`, `Out of Scope Gaps`, boundary coverage if present, and the scoped report verdict.
+Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so the code-review skill can read spec, plan, context snapshot, implementation log, and current diff artifacts. Do not require or read a prior `/prizmkit-test` report here; test generation and test-fix looping happen in the next phase.
 
 Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`.
 

package/bundled/dev-pipeline/templates/sections/phase-review-full.md

@@ -1,8 +1,6 @@
 ### Review — Code Review
 
-Precondition: the `prizmkit-test` checkpoint is completed and `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt` points to the scoped test report.
-
-Before invoking review, read `.prizmkit/specs/{{FEATURE_SLUG}}/test-report-path.txt`, read the referenced `/prizmkit-test` report, and confirm the summary in `context-snapshot.md` under `## PrizmKit Test Gate` matches it. The code-review skill must receive this context through `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` and the Reviewer must assess generated/updated test quality, `In-Scope Failures`, `Baseline Failures`, `Out of Scope Gaps`, boundary coverage if present, and the scoped report verdict.
+Review the completed implementation before the scoped feature test gate runs. Use `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/` so the code-review skill can read spec, plan, context snapshot, implementation log, and current diff artifacts. Do not require or read a prior `/prizmkit-test` report here; test generation and test-fix looping happen in the next phase.
 
 Run `/prizmkit-code-review` with `artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/`.
 

package/bundled/dev-pipeline/templates/sections/phase0-test-baseline.md

@@ -1,15 +1,10 @@
-### Phase 0: Record Test Baseline & Detect Test Commands
+### Phase 0: Detect Test Commands
 
 **Step 1 — Detect test commands**: You know this project's tech stack. Identify ALL test commands that apply (e.g., `go test ./...`, `npm test`, `cargo test`, `pytest`, `make test`, etc.). Record them as `TEST_CMDS`.
 
-**Step 2 — Record pre-existing failure baseline**:
-```bash
-# Run each test command, capture output
-($TEST_CMD) 2>&1 | tee /tmp/test-baseline.txt | tail -20
-```
-Save the list of **pre-existing failing tests** (if any) as `BASELINE_FAILURES`. These are known failures that existed before this session — Dev must NOT be blamed for them, but must list them in COMPLETION_SIGNAL.
+**Step 2 — Do not run tests yet**: This phase only records the commands that will be used by the post-review PrizmKit Test gate. Do not run baseline, periodic, or full-suite tests here; test execution happens after code review.
 
-> **Test Output Rule**: Always capture test output to a temp file (`tee /tmp/test-out.txt`). Then grep the file instead of re-running the suite.
+> **Test Output Rule**: When the post-review gate runs, capture test output to a temp file (`tee /tmp/test-out.txt`). Then grep the file instead of re-running the suite.
 
 
 **Checkpoint update**: Run the update script to set step `test-baseline` to `"completed"`:

package/bundled/dev-pipeline/templates/sections/test-failure-recovery-agent.md

@@ -1,6 +1,6 @@
 ## Test Failure Recovery Protocol
 
-Use this protocol whenever implementation or review tests fail. Its purpose is to distinguish tolerated pre-existing failures from blockers introduced by this session.
+Use this protocol whenever the post-review PrizmKit Test gate fails. Its purpose is to distinguish tolerated pre-existing failures from blockers introduced by this session.
 
 ### Failure Classes
 
@@ -25,7 +25,7 @@ After **6 total test-fix rounds**, STOP trial-and-error. Write failure-log listi
 
 ### Success Rule
 
-Proceed to review only when:
+Proceed to retrospective/commit only when:
 
 1. all new regressions are fixed;
 2. baseline failures are documented;

package/bundled/dev-pipeline/templates/sections/test-failure-recovery-lite.md

@@ -1,6 +1,6 @@
 ## Test Failure Recovery Protocol
 
-Use this protocol whenever implementation tests fail. Its purpose is to distinguish tolerated pre-existing failures from blockers introduced by this session.
+Use this protocol whenever the post-review PrizmKit Test gate fails. Its purpose is to distinguish tolerated pre-existing failures from blockers introduced by this session.
 
 ### Failure Classes
 

package/bundled/dev-pipeline/tests/test_generate_bootstrap_prompt.py

@@ -406,7 +406,7 @@ class TestProcessModeBlocks:
 # ---------------------------------------------------------------------------
 
 class TestScopedFeatureTestGate:
-    def test_standard_sections_insert_prizmkit_test_between_implement_and_review(self):
+    def test_standard_sections_insert_prizmkit_test_after_review(self):
         sections_dir = Path(__file__).resolve().parents[1] / "templates" / "sections"
         sections = assemble_sections(
             "standard",
@@ -419,7 +419,7 @@ class TestScopedFeatureTestGate:
         names = [name for name, _ in sections]
 
         assert "phase-prizmkit-test" in names
-        assert names.index("phase-implement") < names.index("phase-prizmkit-test") < names.index("phase-review")
+        assert names.index("phase-implement") < names.index("phase-review") < names.index("phase-prizmkit-test")
 
         rendered = "\n".join(content for _, content in sections)
         assert "prizmkit-test-gate.py" in rendered
@@ -430,8 +430,12 @@ class TestScopedFeatureTestGate:
         assert ".prizmkit/refactor/" in rendered  # explicitly forbidden in the gate wording
         assert "latest_test_report=$(python3" not in rendered
         assert "boundary_columns_ok=$(awk" not in rendered
+        assert "Runs tests using `TEST_CMD` after each task" not in rendered
+        assert "Run the full test suite" not in rendered
+        assert "before code review" not in rendered
+        assert "Precondition: the `prizmkit-test` checkpoint is completed" not in rendered
 
-    def test_lite_sections_include_prizmkit_test_before_commit(self):
+    def test_lite_sections_include_review_before_prizmkit_test_and_commit(self):
         sections_dir = Path(__file__).resolve().parents[1] / "templates" / "sections"
         sections = assemble_sections(
             "lite",
@@ -444,10 +448,10 @@ class TestScopedFeatureTestGate:
         names = [name for name, _ in sections]
 
         assert "phase-prizmkit-test" in names
-        assert "phase-review" not in names
-        assert names.index("phase-implement") < names.index("phase-prizmkit-test") < names.index("phase-commit")
+        assert "phase-review" in names
+        assert names.index("phase-implement") < names.index("phase-review") < names.index("phase-prizmkit-test") < names.index("phase-commit")
 
-    def test_checkpoint_orders_prizmkit_test_between_implement_and_review(self):
+    def test_checkpoint_orders_prizmkit_test_after_review(self):
         sections_dir = Path(__file__).resolve().parents[1] / "templates" / "sections"
         sections = assemble_sections(
             "standard",
@@ -468,7 +472,7 @@ class TestScopedFeatureTestGate:
         )
         skills = [step["skill"] for step in checkpoint["steps"]]
 
-        assert skills.index("prizmkit-implement") < skills.index("prizmkit-test") < skills.index("prizmkit-code-review")
+        assert skills.index("prizmkit-implement") < skills.index("prizmkit-code-review") < skills.index("prizmkit-test")
         test_step = next(step for step in checkpoint["steps"] if step["skill"] == "prizmkit-test")
         assert test_step["required_artifacts"] == [
             ".prizmkit/specs/123-scope-test/test-report-path.txt",

package/bundled/dev-pipeline/tests/test_python_runner_parity.py

@@ -291,7 +291,7 @@ def test_runner_models_parse_legacy_args_and_environment(tmp_path):
     invocation = parse_invocation(
         family,
         "run",
-        ("custom-list.json", "F-2", "--features", "F-003:F-001", "--max-infra-retries=5", "--mode", "full", "--critic"),
+        ("custom-list.json", "F-2", "--features", "F-003:F-001", "--max-retries", "7", "--max-infra-retries=5", "--mode", "full", "--critic"),
     )
     env = RunnerEnvironment.from_env(
         {
@@ -300,6 +300,7 @@ def test_runner_models_parse_legacy_args_and_environment(tmp_path):
             "DEV_BRANCH": "dev/custom",
             "VERBOSE": "1",
             "STALE_KILL_THRESHOLD": "9",
+            "MAX_RETRIES": "9",
             "STRICT_BEHAVIOR_CHECK": "false",
         }
     )
@@ -307,6 +308,7 @@ def test_runner_models_parse_legacy_args_and_environment(tmp_path):
     assert invocation.list_path == Path("custom-list.json")
     assert invocation.item_id == "F-002"
     assert invocation.item_filter == "F-003:F-001"
+    assert invocation.max_retries == 7
     assert invocation.max_infra_retries == 5
     assert invocation.mode == "full"
     assert invocation.critic is True
@@ -314,6 +316,7 @@ def test_runner_models_parse_legacy_args_and_environment(tmp_path):
     assert env.auto_push is True
     assert env.dev_branch == "dev/custom"
     assert env.stale_kill_threshold_seconds == 9
+    assert env.max_retries == 9
     assert env.strict_behavior_check is False
 
 
@@ -355,6 +358,91 @@ def test_status_bridge_preserves_feature_filter_selection(tmp_path):
     assert update.data["context_overflow_count"] == 1
 
 
+def test_status_bridge_passes_max_retries_from_invocation_or_environment(monkeypatch, tmp_path):
+    import subprocess
+    from prizmkit_runtime.runner_models import family_for, parse_invocation
+    from prizmkit_runtime.runner_status import status_text
+
+    paths = _make_paths(tmp_path)
+    family = family_for("feature", paths)
+    explicit = parse_invocation(family, "run", ("--max-retries", "8"))
+    explicit = explicit.__class__(**{**explicit.__dict__, "list_path": paths.feature_plan})
+    fallback = parse_invocation(family, "run", ())
+    fallback = fallback.__class__(**{**fallback.__dict__, "list_path": paths.feature_plan})
+    calls = []
+
+    def fake_run(command, **kwargs):
+        calls.append(tuple(command))
+        return subprocess.CompletedProcess(command, 0, stdout="PIPELINE_COMPLETE\n", stderr="")
+
+    monkeypatch.setattr(subprocess, "run", fake_run)
+    monkeypatch.setenv("MAX_RETRIES", "6")
+
+    status_text(family, explicit, paths.project_root)
+    status_text(family, fallback, paths.project_root)
+
+    first = calls[0]
+    second = calls[1]
+    assert first[first.index("--max-retries") + 1] == "8"
+    assert second[second.index("--max-retries") + 1] == "6"
+
+
+def test_stop_on_failure_does_not_stop_on_retryable_pending(monkeypatch, tmp_path):
+    from prizmkit_runtime import runners
+    from prizmkit_runtime.runner_models import family_for, parse_invocation
+
+    paths = _make_paths(tmp_path)
+    family = family_for("feature", paths)
+    invocation = parse_invocation(family, "run", ())
+    invocation = invocation.__class__(**{**invocation.__dict__, "list_path": paths.feature_plan})
+    queue = [
+        SimpleNamespace(ok=True, stdout=json.dumps({"feature_id": "F-001"}), data={"feature_id": "F-001"}, text=""),
+        SimpleNamespace(ok=True, stdout=json.dumps({"feature_id": "F-002"}), data={"feature_id": "F-002"}, text=""),
+        SimpleNamespace(ok=True, stdout="PIPELINE_COMPLETE\n", data=None, text="PIPELINE_COMPLETE"),
+    ]
+    processed = []
+
+    monkeypatch.setenv("STOP_ON_FAILURE", "1")
+    monkeypatch.setattr(runners, "get_next", lambda *args, **kwargs: queue.pop(0))
+
+    def fake_process(_family, _invocation, item_id, _paths, *, initial_metadata):
+        processed.append(item_id)
+        return "pending" if item_id == "F-001" else "completed"
+
+    monkeypatch.setattr(runners, "_process_item", fake_process)
+
+    result = runners._run_pipeline(family, invocation, paths)
+
+    assert processed == ["F-001", "F-002"]
+    assert result.completed is True
+    assert result.stopped_reason == "pipeline_complete"
+    assert result.details == ("F-001:pending", "F-002:completed")
+
+
+def test_stop_on_failure_still_stops_on_terminal_status(monkeypatch, tmp_path):
+    from prizmkit_runtime import runners
+    from prizmkit_runtime.runner_models import family_for, parse_invocation
+
+    paths = _make_paths(tmp_path)
+    family = family_for("feature", paths)
+    invocation = parse_invocation(family, "run", ())
+    invocation = invocation.__class__(**{**invocation.__dict__, "list_path": paths.feature_plan})
+
+    monkeypatch.setenv("STOP_ON_FAILURE", "1")
+    monkeypatch.setattr(
+        runners,
+        "get_next",
+        lambda *args, **kwargs: SimpleNamespace(ok=True, stdout=json.dumps({"feature_id": "F-001"}), data={"feature_id": "F-001"}, text=""),
+    )
+    monkeypatch.setattr(runners, "_process_item", lambda *args, **kwargs: "failed")
+
+    result = runners._run_pipeline(family, invocation, paths)
+
+    assert result.completed is False
+    assert result.stopped_reason == "stop_on_failure"
+    assert result.last_status == "failed"
+
+
 def test_prompt_generation_result_preserves_per_item_model(monkeypatch, tmp_path):
     from prizmkit_runtime.runner_models import SessionPaths, family_for, parse_invocation
     from prizmkit_runtime.runner_prompts import generate_prompt

package/bundled/skills/_metadata.json

@@ -1,5 +1,5 @@
 {
-  "version": "1.1.101",
+  "version": "1.1.102",
   "skills": {
     "prizm-kit": {
       "description": "Full-lifecycle dev toolkit. Covers spec-driven development, Prizm context docs, code quality, debugging, deployment, and knowledge management.",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "prizmkit",
-  "version": "1.1.101",
+  "version": "1.1.102",
   "description": "Create a new PrizmKit-powered project with clean initialization — no framework dev files, just what you need.",
   "type": "module",
   "bin": {

package/src/config.js

@@ -54,7 +54,7 @@ import {
   confirmTeamMode,
   confirmPipeline,
 } from './prompts.js';
-import { expandPlatforms, isKnownPlatform, platformLabel, projectMemoryFile } from './platforms.js';
+import { isKnownPlatform, platformLabel, projectMemoryFile, resolvePayloadPlatforms } from './platforms.js';
 import { normalizeRuntime, runtimeLabel } from './runtimes.js';
 
 const __dirname = dirname(fileURLToPath(import.meta.url));
@@ -490,9 +490,9 @@ export async function runConfig(directory, options = {}) {
   }
 
   // Warn about destructive changes
-  if (newConfig.platform !== currentConfig.platform) {
-    const oldPlatforms = expandPlatforms(currentConfig.platform);
-    const newPlatforms = expandPlatforms(newConfig.platform);
+  if (newConfig.platform !== currentConfig.platform || newConfig.aiCli !== currentConfig.aiCli) {
+    const oldPlatforms = resolvePayloadPlatforms(currentConfig.platform, currentConfig.aiCli);
+    const newPlatforms = resolvePayloadPlatforms(newConfig.platform, newConfig.aiCli);
     const dropping = oldPlatforms.filter(p => !newPlatforms.includes(p));
     if (dropping.length) {
       console.log(chalk.yellow(`\n  ⚠ 将移除 ${dropping.map(platformLabel).join(', ')} 的 PrizmKit 文件`));
@@ -521,8 +521,8 @@ export async function runConfig(directory, options = {}) {
 
   console.log(chalk.bold('\n  应用配置变更...\n'));
 
-  const oldPlatforms = expandPlatforms(currentConfig.platform);
-  const newPlatforms = expandPlatforms(newConfig.platform);
+  const oldPlatforms = resolvePayloadPlatforms(currentConfig.platform, currentConfig.aiCli);
+  const newPlatforms = resolvePayloadPlatforms(newConfig.platform, newConfig.aiCli);
   const platformsToRemove = oldPlatforms.filter(p => !newPlatforms.includes(p));
   const platformsToAdd = newPlatforms.filter(p => !oldPlatforms.includes(p));
   const platformsToUpdate = newPlatforms.filter(p => oldPlatforms.includes(p));
@@ -554,7 +554,7 @@ export async function runConfig(directory, options = {}) {
       version: pkg.version, platform: newConfig.platform, suite: newConfig.suite,
       skills: newSkillList, agents: newAgentFiles, rules: newRuleFiles,
       pipeline: newPipelineFiles, team: newConfig.team, aiCli: newConfig.aiCli,
-      runtime: newConfig.runtime, rulesPreset: newConfig.rules, extras: oldManifest?.files?.extras || [],
+      runtime: newConfig.runtime, rulesPreset: newConfig.rules, extras: oldManifest?.files?.extras || [], payloadPlatforms: newPlatforms,
     });
     const diff = diffManifest(oldManifest, newTempManifest);
 
@@ -580,6 +580,7 @@ export async function runConfig(directory, options = {}) {
     || newConfig.runtime !== currentConfig.runtime
     || newConfig.suite !== currentConfig.suite
     || newConfig.rules !== currentConfig.rules
+    || newConfig.aiCli !== currentConfig.aiCli
     || newConfig.team !== currentConfig.team;
 
   for (const p of allTargetPlatforms) {
@@ -675,6 +676,7 @@ export async function runConfig(directory, options = {}) {
     team: newConfig.team,
     aiCli: newConfig.aiCli,
     rulesPreset: newConfig.rules,
+    payloadPlatforms: newPlatforms,
     extras: oldManifest?.files?.extras || [],
   });
 

package/src/external-skills.js

@@ -14,7 +14,7 @@
 import { execSync } from 'node:child_process';
 import path from 'path';
 import fs from 'fs-extra';
-import { expandPlatforms } from './platforms.js';
+import { resolvePayloadPlatforms } from './platforms.js';
 
 /**
  * @param {Object} skill        - Skill definition from external_skills.known
@@ -23,9 +23,9 @@ import { expandPlatforms } from './platforms.js';
  *   skill.installAll {boolean} - If true, install all skills from the repo
  * @param {string} platform     - 'claude' | 'codebuddy' | 'codex' | 'all' ('both' supported only for legacy manifests)
  * @param {string} projectRoot  - Target project root
- * @param {boolean} dryRun
+ * @param {string} [aiCli]       - optional AI CLI command that may require an additional payload platform
  */
-export async function installExternalSkill(skill, platform, projectRoot, dryRun) {
+export async function installExternalSkill(skill, platform, projectRoot, dryRun, aiCli = '') {
   if (!skill.repo) {
     throw new Error(`Skill "${skill.name}" has no repo URL defined`);
   }
@@ -40,7 +40,7 @@ export async function installExternalSkill(skill, platform, projectRoot, dryRun)
     if (skill.installAll) {
       console.log(`      [dry-run] → all skills from ${skill.repo}`);
     } else {
-      const targets = getTargetDirs(platform, projectRoot)
+      const targets = getTargetDirs(platform, projectRoot, aiCli)
         .map(d => path.relative(projectRoot, path.join(d, skill.name, 'SKILL.md')));
       for (const t of targets) console.log(`      [dry-run] → ${t}`);
     }
@@ -61,7 +61,7 @@ export async function installExternalSkill(skill, platform, projectRoot, dryRun)
       .filter(d => d.isDirectory())
       .map(d => d.name);
 
-    for (const targetDir of getTargetDirs(platform, projectRoot)) {
+    for (const targetDir of getTargetDirs(platform, projectRoot, aiCli)) {
       for (const skillName of skillDirs) {
         const srcDir = path.join(agentsSkillsDir, skillName);
         const destDir = path.join(targetDir, skillName);
@@ -81,7 +81,7 @@ export async function installExternalSkill(skill, platform, projectRoot, dryRun)
 
     const content = await fs.readFile(canonicalSkillMd, 'utf8');
 
-    for (const targetDir of getTargetDirs(platform, projectRoot)) {
+    for (const targetDir of getTargetDirs(platform, projectRoot, aiCli)) {
       const skillDir = path.join(targetDir, skill.name);
       if (path.resolve(skillDir) === path.resolve(path.dirname(canonicalSkillMd))) {
         continue;
@@ -94,7 +94,7 @@ export async function installExternalSkill(skill, platform, projectRoot, dryRun)
 
   // Clean up npx skills artifacts after copying. Preserve .agents/skills when
   // Codex is selected because it is the official repository skill location.
-  if (!expandPlatforms(platform).includes('codex')) {
+  if (!resolvePayloadPlatforms(platform, aiCli).includes('codex')) {
     await fs.remove(path.join(projectRoot, '.agents')).catch(() => {});
   }
   await fs.remove(path.join(projectRoot, 'skills-lock.json')).catch(() => {});
@@ -104,15 +104,15 @@ export async function installExternalSkill(skill, platform, projectRoot, dryRun)
  * Remove all artifacts created by `npx skills` that are not needed by PrizmKit.
  * Call this once after all external skills have been installed.
  *
- * @param {string} projectRoot
  * @param {string} platform - 'claude' | 'codebuddy' | 'codex' | 'all' ('both' supported only for legacy manifests)
+ * @param {string} [aiCli] - optional AI CLI command that may require an additional payload platform
  */
-export async function cleanExternalSkillArtifacts(projectRoot, platform) {
+export async function cleanExternalSkillArtifacts(projectRoot, platform, aiCli = '') {
   const keepDirs = new Set(
-    getTargetDirs(platform, projectRoot).map(d => path.resolve(d))
+    getTargetDirs(platform, projectRoot, aiCli).map(d => path.resolve(d))
   );
 
-  const selectedPlatforms = expandPlatforms(platform);
+  const selectedPlatforms = resolvePayloadPlatforms(platform, aiCli);
   const dirsToCheck = [
     selectedPlatforms.includes('codex') ? null : '.agents',
     '.agent',
@@ -152,7 +152,7 @@ export async function cleanExternalSkillArtifacts(projectRoot, platform) {
     const abs = path.join(projectRoot, dir);
     if (!await fs.pathExists(abs)) continue;
     // Only remove if this platform dir is entirely unneeded
-    const isKept = getTargetDirs(platform, projectRoot)
+    const isKept = getTargetDirs(platform, projectRoot, aiCli)
       .some(d => path.resolve(d).startsWith(path.resolve(abs) + path.sep));
     if (isKept) continue;
     try {
@@ -166,8 +166,8 @@ export async function cleanExternalSkillArtifacts(projectRoot, platform) {
   await fs.remove(path.join(projectRoot, 'skills-lock.json')).catch(() => {});
 }
 
-function getTargetDirs(platform, projectRoot) {
-  return expandPlatforms(platform).map(p => {
+function getTargetDirs(platform, projectRoot, aiCli = '') {
+  return resolvePayloadPlatforms(platform, aiCli).map(p => {
     if (p === 'codebuddy') return path.join(projectRoot, '.codebuddy', 'skills');
     if (p === 'codex') return path.join(projectRoot, '.agents', 'skills');
     return path.join(projectRoot, '.claude', 'skills');

package/src/manifest.js

@@ -9,7 +9,7 @@
 
 import fs from 'fs-extra';
 import path from 'path';
-import { expandPlatforms, projectMemoryFile } from './platforms.js';
+import { projectMemoryFile, resolvePayloadPlatforms } from './platforms.js';
 import { normalizeRuntime } from './runtimes.js';
 
 const MANIFEST_FILE = 'manifest.json';
@@ -58,7 +58,7 @@ export async function writeManifest(projectRoot, data) {
  * @param {boolean} params.team - whether team config was installed
  * @param {string} [params.aiCli] - AI CLI command
  * @param {string} [params.rulesPreset] - rules preset name
- * @param {string[]} [params.extras] - installed extras (e.g. ['playwright-cli'])
+ * @param {string[]} [params.payloadPlatforms] - concrete platform payloads installed for platform + aiCli
  * @returns {Object} manifest
  */
 export function buildManifest({
@@ -74,8 +74,12 @@ export function buildManifest({
   aiCli,
   rulesPreset,
   extras,
+  payloadPlatforms,
 }) {
   const now = new Date().toISOString();
+  const installedPayloadPlatforms = payloadPlatforms?.length
+    ? [...new Set(payloadPlatforms)]
+    : resolvePayloadPlatforms(platform, aiCli);
   return {
     version,
     installedAt: now,
@@ -95,7 +99,8 @@ export function buildManifest({
       rules: [...rules],
       pipeline: pipeline ? [...pipeline] : [],
       extras: extras ? [...extras] : [],
-      other: [...new Set(expandPlatforms(platform).map(projectMemoryFile).filter(Boolean))],
+      payloadPlatforms: installedPayloadPlatforms,
+      other: [...new Set(installedPayloadPlatforms.map(projectMemoryFile).filter(Boolean))],
     },
   };
 }

package/src/platforms.js

@@ -10,6 +10,31 @@ export function expandPlatforms(platform) {
   return PLATFORM_GROUPS[platform] || LEGACY_PLATFORM_GROUPS[platform] || [platform];
 }
 
+export function platformFromAiCli(aiCli) {
+  if (!aiCli || typeof aiCli !== 'string') return null;
+  const token = aiCli.trim().split(/\s+/)[0]?.replace(/^['"]|['"]$/g, '');
+  const name = token.split(/[\\/]/).pop()?.toLowerCase() || '';
+  if (name.includes('claude')) return 'claude';
+  if (name.includes('codex')) return 'codex';
+  if (name.includes('cbc') || name.includes('codebuddy')) return 'codebuddy';
+  return null;
+}
+
+export function uniquePlatforms(platforms) {
+  return [...new Set(platforms.filter(p => PLATFORM_IDS.includes(p)))];
+}
+
+export function resolvePayloadPlatforms(platform, aiCli = '') {
+  const selected = uniquePlatforms(expandPlatforms(platform));
+  const cliPlatform = platformFromAiCli(aiCli);
+  return uniquePlatforms([...selected, cliPlatform].filter(Boolean));
+}
+
+export function payloadPlatformLabel(platform, aiCli = '') {
+  const payload = resolvePayloadPlatforms(platform, aiCli);
+  return payload.map(platformLabel).join(' + ');
+}
+
 export function isKnownPlatform(platform) {
   return PLATFORM_IDS.includes(platform) || Object.hasOwn(PLATFORM_GROUPS, platform);
 }

package/src/scaffold.js

@@ -27,7 +27,7 @@ import {
 } from './metadata.js';
 import { generateGitignore } from './gitignore-template.js';
 import { buildManifest, writeManifest } from './manifest.js';
-import { defaultCliForPlatform, expandPlatforms, isKnownPlatform, platformLabel, projectMemoryFile } from './platforms.js';
+import { defaultCliForPlatform, expandPlatforms, isKnownPlatform, platformLabel, projectMemoryFile, resolvePayloadPlatforms } from './platforms.js';
 import { normalizeRuntime, runtimeLabel } from './runtimes.js';
 
 const __scaffoldDirname = dirname(fileURLToPath(import.meta.url));
@@ -1436,6 +1436,7 @@ export async function scaffold(config) {
     throw new Error(`Unknown platform "${platform}". Expected codebuddy, claude, codex, or all.`);
   }
   const platforms = expandPlatforms(platform);
+  const payloadPlatforms = resolvePayloadPlatforms(platform, aiCli);
 
   if (dryRun) {
     console.log(chalk.yellow('\n  [DRY RUN] 预览将要安装的内容:\n'));
@@ -1446,8 +1447,8 @@ export async function scaffold(config) {
 
   console.log('');
 
-  for (const p of platforms) {
-    console.log(chalk.bold(`  正在安装 ${platformLabel(p)} 环境...\n`));
+  for (const p of payloadPlatforms) {
+    console.log(chalk.bold(`  正在安装 ${platformLabel(p)} 环境...${platforms.includes(p) ? '' : chalk.gray(' (AI CLI payload)')}\n`));
 
     // 1. Skills
     console.log(chalk.blue('    技能文件:'));
@@ -1515,7 +1516,7 @@ export async function scaffold(config) {
       const skillDef = knownSkills.find(s => s.name === name);
       if (skillDef) {
         try {
-          await installExternalSkill(skillDef, platform, projectRoot, dryRun);
+          await installExternalSkill(skillDef, platform, projectRoot, dryRun, aiCli);
           console.log(chalk.green(`      ✓ ${name} (external)`));
         } catch (e) {
           console.log(chalk.yellow(`      ⚠ ${name} 安装失败，跳过: ${e.message}`));
@@ -1523,7 +1524,7 @@ export async function scaffold(config) {
       }
     }
     if (!dryRun) {
-      await cleanExternalSkillArtifacts(projectRoot, platform);
+      await cleanExternalSkillArtifacts(projectRoot, platform, aiCli);
     }
   }
 
@@ -1545,7 +1546,7 @@ export async function scaffold(config) {
   // 11. Clean up stray .codebuddy/ directory left by third-party tools (e.g. npx skills)
   // when installing for Claude Code only. CodeBuddy files should never appear in a
   // claude-only install.
-  if (!dryRun && !platforms.includes('codebuddy')) {
+  if (!dryRun && !payloadPlatforms.includes('codebuddy')) {
     const strayCbDir = path.join(projectRoot, '.codebuddy');
     if (await fs.pathExists(strayCbDir)) {
       const entries = await fs.readdir(strayCbDir);
@@ -1584,6 +1585,7 @@ export async function scaffold(config) {
       aiCli,
       rulesPreset: rulesPresetName,
       extras: activeExtras,
+      payloadPlatforms,
     });
     await writeManifest(projectRoot, manifest);
     console.log(chalk.green('      ✓ .prizmkit/manifest.json'));
@@ -1632,6 +1634,9 @@ export async function scaffold(config) {
     console.log(chalk.gray(`  安装统计:`));
     console.log(chalk.gray(`    技能: ${suiteLabel}`));
     console.log(chalk.gray(`    平台: ${platforms.map(platformLabel).join(' + ')}`));
+    if (payloadPlatforms.length !== platforms.length || payloadPlatforms.some(p => !platforms.includes(p))) {
+      console.log(chalk.gray(`    AI CLI 载荷平台: ${payloadPlatforms.map(platformLabel).join(' + ')}`));
+    }
     console.log(chalk.gray(`    运行时: ${runtimeLabel(runtime)}`));
     console.log(chalk.gray(`    团队: ${team ? '已启用' : '未启用'}`));
     console.log(chalk.gray(`    流水线: ${pipeline ? '已安装' : '未安装'}`));

package/src/upgrade.js

@@ -27,6 +27,7 @@ import {
   installGitHook,
   installGitignore,
   installProjectMemory,
+  installTeamConfig,
   resolvePipelineFileList,
   findStaleManagedPipelineFiles,
   removeStaleManagedPipelineFiles,
@@ -34,7 +35,7 @@ import {
   resolveSkillList,
   EXTRAS_REGISTRY,
 } from './scaffold.js';
-import { expandPlatforms, platformLabel } from './platforms.js';
+import { platformLabel, resolvePayloadPlatforms } from './platforms.js';
 import { normalizeRuntime, runtimeLabel } from './runtimes.js';
 
 const __dirname = dirname(fileURLToPath(import.meta.url));
@@ -276,6 +277,8 @@ export async function runUpgrade(directory, options = {}) {
     ...(!oldManifest?.files?.extras ? Object.keys(EXTRAS_REGISTRY) : []),
   ])].filter(name => EXTRAS_REGISTRY[name]);  // only keep extras that still exist in registry
 
+  const payloadPlatforms = resolvePayloadPlatforms(platform, aiCli);
+
   // 4. Build new manifest and compute diff
   const newManifest = buildManifest({
     version: pkg.version,
@@ -290,6 +293,7 @@ export async function runUpgrade(directory, options = {}) {
     aiCli,
     rulesPreset,
     extras: extrasToInstall,
+    payloadPlatforms,
   });
 
   // Preserve original installedAt
@@ -349,7 +353,7 @@ export async function runUpgrade(directory, options = {}) {
   }
 
   // 7. Execute
-  const platforms = expandPlatforms(platform);
+  const platforms = payloadPlatforms;
 
   // 7a. Remove orphaned files
   if (diff.skills.removed.length || diff.agents.removed.length || diff.rules.removed.length || diff.pipeline.removed.length || staleManagedPipelineFiles.length) {
@@ -396,6 +400,11 @@ export async function runUpgrade(directory, options = {}) {
     console.log(chalk.blue('\n    Settings & Rules:'));
     await installSettings(p, projectRoot, { pipeline, rules: rulesPreset }, dryRun, runtime);
 
+    if (team) {
+      console.log(chalk.blue('\n    Team Config:'));
+      await installTeamConfig(p, projectRoot, dryRun);
+    }
+
     console.log('');
   }