prizmkit 1.1.194 → 1.1.196

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,5 +1,5 @@
1
1
  {
2
- "frameworkVersion": "1.1.194",
3
- "bundledAt": "2026-11-21T12:21:51.513Z",
4
- "bundledFrom": "75fa3a8"
2
+ "frameworkVersion": "1.1.196",
3
+ "bundledAt": "2026-11-21T15:21:55.168Z",
4
+ "bundledFrom": "f0c45c2"
5
5
  }
@@ -660,6 +660,8 @@ AI_CLI=agent python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
660
660
  AI_CLI=agy python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
661
661
  ```
662
662
 
663
+ `diagnostics ai-cli` emits structured resolution and capability fields (`detected_cli`, `platform`, `source`, `runtime_mode`, `model_arg_supported`, `effort_flag_mode` as `flag` | `config` | `none` | `custom_ignored`, `supported_effort_values`, and `cli_available`, plus optional `model` and `note` lines). It exits non-zero when no AI CLI can be resolved (`detected_cli: (not found)`), and reports `cli_available: false` with a diagnostic `note:` when an explicitly configured CLI executable is not found on `PATH`.
664
+
663
665
  ### Cursor Agent native headless contract
664
666
 
665
667
  Install Cursor Agent separately (`curl https://cursor.com/install -fsS | bash` on macOS/Linux/WSL, or `irm 'https://cursor.com/install?win32=true' | iex` in native Windows PowerShell). Verify and authenticate with `agent --version`, `agent login`, and `agent status`; update with `agent update`. Because `agent` is a generic executable name, PATH fallback verifies Cursor identity and the required help flags before selecting it.
@@ -670,7 +672,7 @@ For a Pipeline session, PrizmKit runs:
670
672
  agent --print --force --trust --output-format stream-json [--model <model>]
671
673
  ```
672
674
 
673
- The complete coordinator Prompt is sent through stdin and the active task checkout is the process `cwd`. PrizmKit does not pass Cursor `--worktree`, does not disable Cursor sandbox policy, and does not create or modify `.cursor/cli.json`; user and enterprise deny policy continues to apply. Authentication is inherited from `agent login` or `CURSOR_API_KEY`, never serialized into PrizmKit config, argv, or logs. Cursor exposes no documented effort parameter, so `PRIZMKIT_EFFORT` must remain unset for `AI_CLI=agent`; a non-empty value fails with an actionable diagnostic.
675
+ The complete coordinator Prompt is sent through stdin and the active task checkout is the process `cwd`. PrizmKit does not pass Cursor `--worktree`, does not disable Cursor sandbox policy, and does not create or modify `.cursor/cli.json`; user and enterprise deny policy continues to apply. Authentication is inherited from `agent login` or `CURSOR_API_KEY`, never serialized into PrizmKit config, argv, or logs. Cursor exposes no documented effort parameter, so `PRIZMKIT_EFFORT` must remain unset for `AI_CLI=agent` (a non-empty value fails with an actionable diagnostic); instead, Cursor supports parameterized `--model` identifiers such as `MODEL='claude-opus-4-8[context=1m,effort=high,fast=false]'`.
674
676
 
675
677
  ### Antigravity CLI native headless contract
676
678
 
@@ -2,7 +2,10 @@
2
2
 
3
3
  from __future__ import annotations
4
4
 
5
+ import os
6
+ import shutil
5
7
  from dataclasses import dataclass
8
+ from pathlib import Path
6
9
 
7
10
  from .compat import MIN_PYTHON
8
11
  from .interoperability import UTILITY_ENTRYPOINTS, describe_existing_entrypoint
@@ -303,22 +306,59 @@ def diagnostics_version() -> CommandResult:
303
306
  )
304
307
 
305
308
 
309
+ def _is_cli_executable_available(
310
+ command: str | None,
311
+ path_entries: tuple[Path, ...] | None,
312
+ ) -> bool:
313
+ """Return whether the resolved AI CLI executable exists and is runnable."""
314
+ if not command:
315
+ return False
316
+ if os.sep in command or (os.altsep and os.altsep in command):
317
+ candidate = Path(command)
318
+ if os.name == "nt":
319
+ return candidate.is_file()
320
+ return candidate.is_file() and os.access(command, os.X_OK)
321
+ search_path = (
322
+ os.pathsep.join(str(entry) for entry in path_entries)
323
+ if path_entries is not None
324
+ else os.environ.get("PATH", "")
325
+ )
326
+ return shutil.which(command, path=search_path) is not None
327
+
328
+
306
329
  def diagnostics_ai_cli(paths: RuntimePaths | None = None) -> CommandResult:
307
330
  """Report the family-neutral AI CLI resolution used by runtime commands."""
308
- from .config import load_runtime_config
331
+ from .config import ai_cli_capabilities, load_runtime_config
309
332
 
310
333
  runtime_paths = paths or resolve_runtime_paths()
311
334
  config = load_runtime_config(runtime_paths)
312
335
  cli = config.ai_client.command or "(not found)"
336
+ capabilities = ai_cli_capabilities(config.ai_client.platform)
337
+ supported_efforts = (
338
+ ",".join(capabilities.supported_effort_values)
339
+ if capabilities.supported_effort_values
340
+ else "none"
341
+ )
342
+ env_obj = getattr(config, "environment", None)
343
+ path_entries = getattr(env_obj, "path_entries", None)
344
+ cli_available = _is_cli_executable_available(config.ai_client.command, path_entries)
313
345
  details = [
314
346
  f"detected_cli: {cli}",
315
347
  f"platform: {config.ai_client.platform or 'unresolved'}",
316
348
  f"source: {config.ai_client.source}",
317
349
  f"runtime_mode: {config.runtime_mode}",
350
+ f"model_arg_supported: {'true' if capabilities.model_arg_supported else 'false'}",
351
+ f"effort_flag_mode: {capabilities.effort_flag_mode}",
352
+ f"supported_effort_values: {supported_efforts}",
353
+ f"cli_available: {'true' if cli_available else 'false'}",
318
354
  ]
319
355
  if config.model:
320
356
  details.append(f"model: {config.model}")
321
357
  details.extend(f"note: {note}" for note in config.ai_client.notes)
358
+ if config.ai_client.command and not cli_available:
359
+ details.append(
360
+ f"note: Resolved AI CLI '{cli}' was not found on PATH; install it or adjust PATH/AI_CLI before launching."
361
+ )
322
362
  return CommandResult(
323
363
  0 if config.ai_client.command else 1,
324
364
  "Python runtime AI CLI diagnostics",
@@ -48,6 +48,15 @@ class AIClientResolution:
48
48
  launch_profile: HeadlessLaunchProfile | None = None
49
49
 
50
50
 
51
+ @dataclass(frozen=True)
52
+ class AIClientCapabilities:
53
+ """CLI model-argument and reasoning-effort capabilities for one platform profile."""
54
+
55
+ model_arg_supported: bool
56
+ effort_flag_mode: str
57
+ supported_effort_values: tuple[str, ...]
58
+
59
+
51
60
  @dataclass(frozen=True)
52
61
  class EffortValidationResult:
53
62
  """Platform-aware validation result for PRIZMKIT_EFFORT."""
@@ -92,13 +101,57 @@ AI_CLIENT_PRECEDENCE = (
92
101
  "PATH:agy",
93
102
  )
94
103
 
95
- _EFFORT_VALUES_BY_PLATFORM = {
96
- "claude": ("low", "medium", "high", "xhigh", "max"),
97
- "codex": ("low", "medium", "high", "xhigh"),
98
- "codebuddy": ("low", "medium", "high", "xhigh"),
99
- "pi": ("off", "minimal", "low", "medium", "high", "xhigh", "max"),
100
- "cursor": (),
101
- "antigravity": ("low", "medium", "high", "xhigh", "max"),
104
+ _CAPABILITIES_BY_PLATFORM: dict[str, AIClientCapabilities] = {
105
+ "claude": AIClientCapabilities(
106
+ model_arg_supported=True,
107
+ effort_flag_mode="flag",
108
+ supported_effort_values=("low", "medium", "high", "xhigh", "max"),
109
+ ),
110
+ "codebuddy": AIClientCapabilities(
111
+ model_arg_supported=True,
112
+ effort_flag_mode="flag",
113
+ supported_effort_values=("minimal", "low", "medium", "high", "xhigh", "max"),
114
+ ),
115
+ "antigravity": AIClientCapabilities(
116
+ model_arg_supported=True,
117
+ effort_flag_mode="flag",
118
+ supported_effort_values=("low", "medium", "high", "xhigh", "max"),
119
+ ),
120
+ "pi": AIClientCapabilities(
121
+ model_arg_supported=True,
122
+ effort_flag_mode="flag",
123
+ supported_effort_values=("off", "minimal", "low", "medium", "high", "xhigh", "max"),
124
+ ),
125
+ "codex": AIClientCapabilities(
126
+ model_arg_supported=True,
127
+ effort_flag_mode="config",
128
+ supported_effort_values=("low", "medium", "high", "xhigh"),
129
+ ),
130
+ "cursor": AIClientCapabilities(
131
+ model_arg_supported=True,
132
+ effort_flag_mode="none",
133
+ supported_effort_values=(),
134
+ ),
135
+ "custom": AIClientCapabilities(
136
+ model_arg_supported=False,
137
+ effort_flag_mode="custom_ignored",
138
+ supported_effort_values=(),
139
+ ),
140
+ }
141
+
142
+ _UNRESOLVED_CAPABILITIES = AIClientCapabilities(
143
+ model_arg_supported=False,
144
+ effort_flag_mode="none",
145
+ supported_effort_values=(),
146
+ )
147
+
148
+ _PLATFORM_LABELS: dict[str, str] = {
149
+ "claude": "Claude Code",
150
+ "codex": "Codex",
151
+ "codebuddy": "CodeBuddy",
152
+ "pi": "Pi",
153
+ "cursor": "Cursor Agent CLI",
154
+ "antigravity": "Antigravity",
102
155
  }
103
156
 
104
157
 
@@ -457,52 +510,76 @@ def _resolve_ai_client(paths: RuntimePaths, env: Mapping[str, str]) -> AIClientR
457
510
  )
458
511
 
459
512
 
513
+ def _capability_platform(platform: str | None) -> str | None:
514
+ normalized = _normalize_platform(platform)
515
+ return normalized if normalized in _CAPABILITIES_BY_PLATFORM else None
516
+
517
+
518
+ def ai_cli_capabilities(platform: str | None) -> AIClientCapabilities:
519
+ """Return immutable model and effort capabilities for one platform profile."""
520
+ resolved = _capability_platform(platform)
521
+ if resolved is None:
522
+ return _UNRESOLVED_CAPABILITIES
523
+ return _CAPABILITIES_BY_PLATFORM[resolved]
524
+
525
+
460
526
  def supported_effort_values(platform: str | None) -> tuple[str, ...]:
461
527
  """Return supported PRIZMKIT_EFFORT values for the detected platform."""
462
- return _EFFORT_VALUES_BY_PLATFORM.get(
463
- _concrete_platform(platform) or "codebuddy",
464
- _EFFORT_VALUES_BY_PLATFORM["codebuddy"],
465
- )
528
+ return ai_cli_capabilities(platform).supported_effort_values
466
529
 
467
530
 
468
- def validate_effort(effort: str | None, platform: str | None) -> EffortValidationResult:
469
- """Validate PRIZMKIT_EFFORT without exiting the process."""
470
- supported = supported_effort_values(platform)
531
+ def validate_effort(effort: str | None, *, platform: str | None = None) -> str | None:
532
+ """Validate and normalize PRIZMKIT_EFFORT against platform capabilities."""
471
533
  if effort is None or str(effort).strip() == "":
534
+ return None
535
+ normalized_effort = str(effort).strip().lower()
536
+ resolved_platform = _capability_platform(platform)
537
+ capabilities = ai_cli_capabilities(resolved_platform)
538
+ if normalized_effort in capabilities.supported_effort_values:
539
+ return normalized_effort
540
+ if resolved_platform == "cursor":
541
+ raise ValueError(
542
+ f"PRIZMKIT_EFFORT='{normalized_effort}' is not supported by the detected CLI "
543
+ "(Cursor Agent CLI). Cursor Agent CLI exposes no documented effort parameter; "
544
+ "leave PRIZMKIT_EFFORT unset."
545
+ )
546
+ if resolved_platform == "custom":
547
+ raise ValueError(
548
+ f"PRIZMKIT_EFFORT='{normalized_effort}' is not supported by the custom AI CLI launch profile "
549
+ "(custom). Custom launch profiles do not forward model or effort flags; "
550
+ "leave PRIZMKIT_EFFORT unset or pass provider flags via ai_cli_launch.extra_args."
551
+ )
552
+ if resolved_platform is None:
553
+ raise ValueError(
554
+ f"PRIZMKIT_EFFORT='{normalized_effort}' cannot be validated because no supported AI CLI platform "
555
+ "was resolved. Configure or install a supported AI CLI platform, or leave PRIZMKIT_EFFORT unset."
556
+ )
557
+ label = _PLATFORM_LABELS[resolved_platform]
558
+ raise ValueError(
559
+ f"PRIZMKIT_EFFORT='{normalized_effort}' is not supported by the detected CLI "
560
+ f"({label}). Supported values for {label}: {', '.join(capabilities.supported_effort_values)}"
561
+ )
562
+
563
+
564
+ def _evaluate_effort_config(effort: str | None, platform: str | None) -> EffortValidationResult:
565
+ resolved_platform = _capability_platform(platform)
566
+ supported = ai_cli_capabilities(resolved_platform).supported_effort_values
567
+ try:
568
+ validated = validate_effort(effort, platform=resolved_platform)
569
+ except ValueError as exc:
570
+ normalized = str(effort).strip().lower() if effort is not None else ""
472
571
  return EffortValidationResult(
473
- effort=None,
474
- platform=_concrete_platform(platform),
475
- valid=True,
572
+ effort=normalized or None,
573
+ platform=resolved_platform,
574
+ valid=False,
476
575
  supported_values=supported,
576
+ message=str(exc),
477
577
  )
478
- normalized_effort = str(effort).strip()
479
- valid = normalized_effort in supported
480
- message = ""
481
- if not valid:
482
- label = {
483
- "claude": "Claude Code",
484
- "codex": "Codex",
485
- "codebuddy": "CodeBuddy",
486
- "pi": "Pi",
487
- "cursor": "Cursor Agent CLI",
488
- "antigravity": "Antigravity",
489
- }.get(_concrete_platform(platform) or "codebuddy", "CodeBuddy")
490
- if not supported:
491
- message = (
492
- f"PRIZMKIT_EFFORT='{normalized_effort}' is not supported by the detected CLI "
493
- f"({label}). Cursor Agent CLI exposes no documented effort parameter; leave PRIZMKIT_EFFORT unset."
494
- )
495
- else:
496
- message = (
497
- f"PRIZMKIT_EFFORT='{normalized_effort}' is not supported by the detected CLI "
498
- f"({label}). Supported values for {label}: {', '.join(supported)}"
499
- )
500
578
  return EffortValidationResult(
501
- effort=normalized_effort,
502
- platform=_concrete_platform(platform),
503
- valid=valid,
579
+ effort=validated,
580
+ platform=resolved_platform,
581
+ valid=True,
504
582
  supported_values=supported,
505
- message=message,
506
583
  )
507
584
 
508
585
 
@@ -512,7 +589,7 @@ def load_runtime_config(paths: RuntimePaths, env: Mapping[str, str] | None = Non
512
589
  runtime_env = load_env_file(paths.prizmkit_dir / ".env", provided_env)
513
590
  ai_client = _resolve_ai_client(paths, runtime_env)
514
591
  effort = runtime_env.get("PRIZMKIT_EFFORT")
515
- effort_validation = validate_effort(effort, ai_client.platform)
592
+ effort_validation = _evaluate_effort_config(effort, ai_client.platform)
516
593
  environment = RuntimeEnvironment(
517
594
  env_file=paths.prizmkit_dir / ".env",
518
595
  config_file=paths.prizmkit_dir / "config.json",
@@ -463,20 +463,31 @@ def _payload_command(family: RunnerFamily | _DaemonTarget, paths, list_path: Pat
463
463
  return command
464
464
 
465
465
 
466
+ def _strip_matching_outer_quotes(text: str) -> str:
467
+ if len(text) >= 2 and text[0] == text[-1] and text[0] in "'\"":
468
+ return text[1:-1]
469
+ return text
470
+
471
+
466
472
  def _daemon_env(options: DaemonStartOptions) -> dict[str, str]:
467
473
  env: dict[str, str] = {"PRIZMKIT_LIVE_OUTPUT": "0"}
468
474
  if not options.env_overrides.strip():
469
475
  return env
476
+ posix = os.name != "nt"
470
477
  try:
471
- tokens = shlex.split(options.env_overrides, posix=(os.name != "nt"))
478
+ tokens = shlex.split(options.env_overrides, posix=posix)
472
479
  except ValueError as exc:
473
480
  raise ValueError(f"Invalid --env syntax: {exc}") from exc
474
- for token in tokens:
481
+ for raw_token in tokens:
482
+ token = _strip_matching_outer_quotes(raw_token) if not posix else raw_token
483
+ token_unquoted = token != raw_token
475
484
  if "=" not in token:
476
- raise ValueError(f"Invalid --env token (expected KEY=VALUE): {token}")
485
+ raise ValueError(f"Invalid --env token (expected KEY=VALUE): {raw_token}")
477
486
  key, value = token.split("=", 1)
478
487
  if not _ENV_KEY_PATTERN.fullmatch(key):
479
488
  raise ValueError(f"Invalid --env variable name: {key!r}")
489
+ if not posix and not token_unquoted:
490
+ value = _strip_matching_outer_quotes(value)
480
491
  env[key] = value
481
492
  return env
482
493
 
@@ -1,5 +1,5 @@
1
1
  {
2
- "version": "1.1.194",
2
+ "version": "1.1.196",
3
3
  "skills": {
4
4
  "prizmkit": {
5
5
  "description": "Framework introduction and navigation for the formal single-requirement lifecycle, project initialization, Prizm docs, and independent deployment.",
@@ -111,7 +111,7 @@
111
111
  "hasScripts": true
112
112
  },
113
113
  "feature-pipeline-launcher": {
114
- "description": "Build feature pipeline start, status, logs, stop, and safe retry commands from a validated feature-list.json without executing them.",
114
+ "description": "Build feature pipeline start, status, logs, stop, and safe retry commands from a validated feature-list.json with optional on-demand read-only AI CLI capability probing, without executing pipelines.",
115
115
  "tier": "companion",
116
116
  "category": "orchestration-skill",
117
117
  "subcategory": "pipelines",
@@ -119,7 +119,7 @@
119
119
  "hasScripts": false
120
120
  },
121
121
  "bugfix-pipeline-launcher": {
122
- "description": "Build bugfix pipeline start, status, logs, stop, and safe retry commands from a validated bug-fix-list.json without executing them.",
122
+ "description": "Build bugfix pipeline start, status, logs, stop, and safe retry commands from a validated bug-fix-list.json with optional on-demand read-only AI CLI capability probing, without executing pipelines.",
123
123
  "tier": "companion",
124
124
  "category": "orchestration-skill",
125
125
  "subcategory": "pipelines",
@@ -127,7 +127,7 @@
127
127
  "hasScripts": false
128
128
  },
129
129
  "iterate-pipeline-launcher": {
130
- "description": "Build autonomous product-iteration start, status, logs, stop, and post-pause resume commands without executing them.",
130
+ "description": "Build autonomous product-iteration start, status, logs, stop, and post-pause resume commands with optional on-demand read-only AI CLI capability probing, without executing pipelines.",
131
131
  "tier": "companion",
132
132
  "category": "orchestration-skill",
133
133
  "subcategory": "pipelines",
@@ -1,20 +1,23 @@
1
1
  ---
2
2
  name: "bugfix-pipeline-launcher"
3
- description: "Build bugfix-pipeline commands from a validated .prizmkit/plans/bug-fix-list.json without executing them. Produces foreground or background-daemon start commands plus status, logs, stop, and safe retry commands. Use whenever the user wants to start fixing a prepared bug list, construct a bugfix command, check bugfix status, view logs, stop a run, retry a bug, or clean retry B-001. Trigger on: 'start fixing bugs', 'run bugfix pipeline', 'bugfix status', 'bugfix logs', 'stop bug fix', 'retry bug', 'retry B-001'."
3
+ description: "Build bugfix-pipeline commands from a validated .prizmkit/plans/bug-fix-list.json without executing the pipeline. Supports bounded read-only AI CLI model/effort capability probing on demand and produces foreground or background-daemon start commands plus status, logs, stop, and safe retry commands. Use whenever the user wants to start fixing a prepared bug list, construct a bugfix command, check bugfix status, view logs, stop a run, retry a bug, or clean retry B-001. Trigger on: 'start fixing bugs', 'run bugfix pipeline', 'bugfix status', 'bugfix logs', 'stop bug fix', 'retry bug', 'retry B-001'."
4
4
  ---
5
5
 
6
6
  # Bugfix Pipeline Command Builder
7
7
 
8
- Build commands for the autonomous bugfix pipeline from within an AI CLI conversation. This skill reads configuration and planning inputs, asks only the choices needed to construct an accurate command, displays the command and its effects, then stops.
8
+ Build commands for the autonomous bugfix pipeline from within an AI CLI conversation. This skill reads configuration and planning inputs, runs bounded read-only AI CLI capability probes only when custom model or reasoning effort configuration is needed, asks only the choices required to construct an accurate command, displays the command and its effects, then stops.
9
9
 
10
- ## Command-Only Boundary
10
+ ## Execution Boundary
11
11
 
12
- This skill:
12
+ This skill separates bounded read-only configuration probing from pipeline execution:
13
13
 
14
- - reads a validated bug list and relevant configuration;
15
- - builds prerequisite-check, start, status, logs, stop, and retry commands;
16
- - explains command effects and destructive options before returning them;
17
- - never executes shell commands, starts processes, follows logs, changes pipeline state, fixes source code, or runs verification.
14
+ - **Allowed (bounded read-only configuration probing only)**:
15
+ - read a validated bug list and relevant configuration;
16
+ - build prerequisite-check, start, status, logs, stop, and retry commands;
17
+ - explain command effects and destructive options before returning them;
18
+ - only when the user selects Advanced config or explicitly mentions/asks about a custom model or reasoning effort: run bounded, side-effect-free read-only probes — `python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli`, check whether the resolved CLI executable exists on `PATH`, and inspect `<cli> --help` (plus `<cli> exec --help` when applicable). Read-only probes must never invoke prompt runs, interactive logins, file writes, Git mutations, or pipeline state changes.
19
+ - **Prohibited (pipeline execution and state mutation)**:
20
+ - never execute pipeline commands (`run`, `daemon start`, `stop`, `reset`), start processes, follow logs, change pipeline or Git state, fix source code, or run verification unless explicitly requested by the user.
18
21
 
19
22
  Do not invoke another workflow or implementation skill from this command builder. The user decides where and when to run the returned command.
20
23
 
@@ -24,7 +27,7 @@ Use this skill for:
24
27
 
25
28
  - constructing a foreground or background-daemon bugfix pipeline command;
26
29
  - constructing status, logs, stop, or retry commands;
27
- - translating user-selected runtime options into environment variables.
30
+ - translating user-selected runtime options (`MODEL`, adaptive `PRIZMKIT_EFFORT`, retries, logging) into environment variables.
28
31
 
29
32
  Do not use it to:
30
33
 
@@ -78,7 +81,9 @@ Use a separate `AskUserQuestion` call with exactly three questions:
78
81
  2. **Max retries** — 3, 1, 5, or a user-provided non-negative integer.
79
82
  3. **Advanced config** — No or Yes.
80
83
 
81
- If Advanced config is Yes, load the advanced question set from `${SKILL_DIR}/references/configuration.md`. Advanced configuration contains only stop-on-failure and reasoning effort.
84
+ If Advanced config is No and the user did not specify or ask about a custom model or reasoning effort, skip read-only CLI probing (`diagnostics ai-cli` and `--help`) and record `Read-only CLI probe: skipped (default launch without custom model/effort)`.
85
+
86
+ If Advanced config is Yes — or if the user specified or asked about a custom model or reasoning effort — load `${SKILL_DIR}/references/configuration.md`, run its on-demand read-only CLI capability probe first, and collect confirmed advanced options (stop-on-failure, `MODEL`, and adaptive `PRIZMKIT_EFFORT`) following its four-branch placement rules.
82
87
 
83
88
  ### Step 4: Build command fragments
84
89
 
@@ -90,7 +95,7 @@ Return this prerequisite-check command separately from the pipeline command:
90
95
  python3 ./.prizmkit/dev-pipeline/cli.py diagnostics version && git --version
91
96
  ```
92
97
 
93
- Also include the configured AI CLI check fragment from the reference. These are returned commands, not checks performed by this skill.
98
+ Also include the configured AI CLI check fragment from the reference so the user can verify prerequisites in their own shell. Even when on-demand read-only probing ran during configuration, pipeline commands are never executed by this skill.
94
99
 
95
100
  ### Step 5: Return the start command
96
101
 
@@ -108,7 +113,7 @@ python3 ./.prizmkit/dev-pipeline/cli.py daemon bugfix start .prizmkit/plans/bug-
108
113
  --env "VERBOSE=1 MAX_RETRIES=3"
109
114
  ```
110
115
 
111
- Return only the template matching the confirmed mode and options. Explain that the command has not been executed, then stop.
116
+ Return only the template matching the confirmed mode and options. Explain that pipeline execution was not performed by this skill, then stop.
112
117
 
113
118
  ## Status Command
114
119
 
@@ -199,6 +204,7 @@ For every request, return:
199
204
  2. Input plan path and selected IDs, when applicable.
200
205
  3. Assumptions and safety warnings.
201
206
  4. Exact command block.
202
- 5. `Command execution: not performed by this skill.`
207
+ 5. `Read-only CLI probe: <diagnostics-and-help-summary | skipped (default launch without custom model/effort)>`
208
+ 6. `Pipeline execution: not performed by this skill; run any command above manually when ready.`
203
209
 
204
- Never report command success, process IDs, log locations, bug completion, verification results, or state changes unless the user later supplies authoritative runtime output for explanation.
210
+ Never report pipeline execution success, process IDs, log locations, bug completion, verification results, or state changes unless the user later supplies authoritative runtime output for explanation.
@@ -1,10 +1,10 @@
1
1
  # Configuration Reference — Bugfix Pipeline Command Builder
2
2
 
3
- Use this reference only to translate confirmed user choices into command fragments. The skill returns commands; it does not execute them.
3
+ Use this reference to run bounded on-demand read-only AI CLI capability probing and translate confirmed user choices into command fragments. This skill constructs and returns pipeline commands; it never executes pipeline runs or mutates state unless explicitly requested by the user.
4
4
 
5
5
  ## Configured AI CLI Prerequisite Command
6
6
 
7
- Runtime AI CLI selection is resolved by the canonical runtime across `AI_CLI`, `.prizmkit/config.json` (`ai_cli`), and installed host configuration in the exact order used by `cli.py run`.
7
+ Runtime AI CLI selection is resolved by the canonical runtime across `AI_CLI`, `.prizmkit/config.json` (`ai_cli` and `ai_cli_launch`), `.prizmkit/manifest.json` (`platform`), and installed host configuration in the exact order used by `cli.py run`.
8
8
 
9
9
  Return this cross-platform check command as part of start-command preparation:
10
10
 
@@ -12,7 +12,16 @@ Return this cross-platform check command as part of start-command preparation:
12
12
  python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
13
13
  ```
14
14
 
15
- `diagnostics ai-cli` reports the resolved CLI, source, and runtime mode, and exits non-zero if no configured AI CLI executable is available. Do not claim the configured executable exists until the user supplies successful command output.
15
+ `diagnostics ai-cli` reports `detected_cli`, `platform`, `source`, `runtime_mode`, `model_arg_supported`, `effort_flag_mode`, `supported_effort_values`, and `cli_available`, and exits non-zero if no AI CLI can be resolved (`detected_cli: (not found)`).
16
+
17
+ ## On-Demand Read-Only CLI Capability Probe
18
+
19
+ - **Fast default launch (skip probing)**: When Advanced config is not selected and the user did not mention or ask about a custom model or reasoning effort, skip all read-only CLI probing (`diagnostics ai-cli` and `--help`) and output `Read-only CLI probe: skipped (default launch without custom model/effort)`.
20
+ - **On-demand probe trigger**: Run the bounded read-only probe sequence only when Advanced config is selected or the user explicitly specifies or asks about `MODEL` or `PRIZMKIT_EFFORT` (including when paired with a per-invocation `AI_CLI=<cli>` override):
21
+ 1. Run `python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli` — or `AI_CLI=<cli> python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli` whenever the user explicitly specifies a per-invocation `AI_CLI=<cli>` override (and preserve `AI_CLI=<cli>` in the returned launch command) — then read the structured capability fields directly: `detected_cli`, `platform`, `source`, `model_arg_supported` (`true`|`false`), `effort_flag_mode` (`flag`|`config`|`none`|`custom_ignored`), `supported_effort_values` (comma-separated values or `none`), and `cli_available` (`true`|`false`). Consume these structured fields directly with zero legacy fallback parsing; if any required capability field is absent, treat the probe as failed under Branch 1.
22
+ 2. If `effort_flag_mode` is `custom_ignored`, route directly to Branch 2 without requiring `<cli> --help`.
23
+ 3. Otherwise, verify that `cli_available` is `true` and the resolved `detected_cli` executable exists on `PATH` (when not `(not found)`), then run `<cli> --help` (and `<cli> exec --help` when the CLI exposes an `exec` subcommand or `effort_flag_mode` is `config`) to inspect live model flags, independent effort/thinking options, or model-embedded intensity syntax.
24
+ 4. Keep every probe strictly read-only and bounded: never invoke prompt runs, interactive logins, file writes, Git mutations, or pipeline commands (`run`, `daemon start`, `stop`, `reset`). Record a concise summary of observed capability facts in `Read-only CLI probe:`.
16
25
 
17
26
  ## Base Configuration Mapping
18
27
 
@@ -24,21 +33,53 @@ python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
24
33
 
25
34
  ## Advanced Configuration Questions
26
35
 
27
- Ask only when the user selected Advanced config.
36
+ Ask only when the user selected Advanced config or explicitly requested custom model/effort configuration. Configure `MODEL` before `PRIZMKIT_EFFORT`:
28
37
 
29
38
  1. **Stop on failure**
30
39
  - Off (default) — continue to other eligible tasks after a task exhausts retries.
31
40
  - On — add `STOP_ON_FAILURE=1`.
32
- 2. **Reasoning effort**
41
+ 2. **Model (`MODEL`)**
42
+ - Default — omit `MODEL` so the resolved AI CLI uses its default model.
43
+ - Explicit model — add `MODEL=<name>` (or single-quoted `MODEL='<model-with-level>'` for parameterized model syntax) when `model_arg_supported` is `true` (or when probing failed without diagnostics confirming `model_arg_supported: false` and the user explicitly confirms a model name).
44
+ - If `diagnostics ai-cli` reports `effort_flag_mode: custom_ignored`, warn clearly that the active custom launch profile ignores both `MODEL` and `PRIZMKIT_EFFORT` (pass provider flags via `.prizmkit/config.json` `ai_cli_launch.extra_args`). If `model_arg_supported: false` is reported for an unresolved/unsupported profile, warn that `MODEL` is not forwarded and leave it unset.
45
+ - **Per-item precedence**: an individual bug's `"model"` field in `.prizmkit/plans/bug-fix-list.json` takes precedence over the global `MODEL` environment variable during task execution.
46
+ 3. **Reasoning effort (`PRIZMKIT_EFFORT`)**
33
47
  - Default — omit `PRIZMKIT_EFFORT`.
34
- - `low`, `medium`, `high`, `xhigh`, or `max` — add `PRIZMKIT_EFFORT=<value>` after confirming the configured AI CLI supports it.
48
+ - Adaptive placement — determine whether and how to emit `PRIZMKIT_EFFORT` strictly from the four ordered placement branches below (never hardcode one universal effort list or CLI allowlist).
49
+
50
+ ## Adaptive Placement Branches (Evaluate 1 → 4 Strictly in Order)
51
+
52
+ Evaluate branches strictly from 1 to 4 using `diagnostics ai-cli` capability fields plus live `--help` output; the first matching branch governs placement. Never branch on hardcoded platform CLI allowlists.
35
53
 
36
- These are the complete advanced choices for this command builder.
54
+ 1. **Branch 1 — Probe failed, CLI missing, or insufficient help output**:
55
+ - Condition: `diagnostics ai-cli` fails or exits non-zero, `detected_cli` is `(not found)` or missing on `PATH`, `cli_available` is `false`, required capability fields (`model_arg_supported`, `effort_flag_mode`, `supported_effort_values`, `cli_available`) are absent, OR (when `effort_flag_mode` is not `custom_ignored`) `<cli> --help` fails or yields insufficient output.
56
+ - Action: Never guess or synthesize `PRIZMKIT_EFFORT`. Omit `PRIZMKIT_EFFORT`, state the exact probe gap in `Read-only CLI probe:` and warnings, and allow an optional explicit `MODEL=<name>` only if diagnostics did not confirm `model_arg_supported: false` (otherwise leave `MODEL` unset).
57
+ 2. **Branch 2 — Model-embedded intensity syntax or `custom_ignored` profile**:
58
+ - Condition: `effort_flag_mode` is `custom_ignored` (routed directly without letting `<cli> --help` failure trigger Branch 1), OR (`model_arg_supported` is `true`, no compatible independent effort mechanism applies — `effort_flag_mode` is `none`, or `supported_effort_values` is `none`, or `effort_flag_mode` is `flag` with no compatible effort/thinking flag in `--help`, excluding compatible `effort_flag_mode: config` — AND live `--help` documents encoding intensity inside the model identifier syntax such as `<model>:<level>` or parameterized model notation).
59
+ - Action: Omit `PRIZMKIT_EFFORT`. Under `custom_ignored`, warn clearly that the active custom launch profile ignores both `MODEL` and `PRIZMKIT_EFFORT` (configure provider flags via `.prizmkit/config.json` `ai_cli_launch.extra_args`), never emit `PRIZMKIT_EFFORT` (the runtime rejects non-empty `PRIZMKIT_EFFORT` pre-launch on custom profiles), and omit `MODEL` unless the user explicitly overrides. Under model-embedded intensity syntax, guide the user to set `MODEL=<model-with-level>` (or leave `MODEL` unset for the CLI default).
60
+ 3. **Branch 3 — Independent effort explicitly unsupported**:
61
+ - Condition: Branches 1 and 2 do not match, AND `effort_flag_mode` is `none`, OR `supported_effort_values` is `none`, OR `effort_flag_mode` is `flag` yet live `--help` shows no compatible effort/thinking option (excluding compatible `effort_flag_mode: config`).
62
+ - Action: Do not emit `PRIZMKIT_EFFORT` (keep it unset so the runtime does not reject it pre-launch). Still permit `MODEL=<name>` when `model_arg_supported` is `true`.
63
+ 4. **Branch 4 — Independent effort compatible**:
64
+ - Condition: (`effort_flag_mode` is `flag` with live `--help` confirming a compatible effort/thinking option, OR `effort_flag_mode` is `config`) AND `supported_effort_values` is non-empty (not `none`).
65
+ - Action: When live `--help` enumerates concrete effort/thinking values, intersect those enumerated choices with `supported_effort_values`; if the intersection is empty, omit `PRIZMKIT_EFFORT` and disclose the capability mismatch rather than emitting an unvalidated value. Otherwise present the valid choices and emit `PRIZMKIT_EFFORT=<value>` (or omit for default). If the confirmed `MODEL` value already embeds a reasoning-effort suffix or parameter (for CLIs that accept both model-embedded intensity and an independent effort flag), omit `PRIZMKIT_EFFORT` by default and note that intensity is already specified in `MODEL` so conflicting intensity settings are not sent together.
37
66
 
38
67
  ## Command Construction Rules
39
68
 
40
- - Foreground: place environment assignments before `python3`.
41
- - Background daemon: place assignments inside one `--env "..."` value.
69
+ - Foreground: place environment assignments (`AI_CLI=...` when overridden, `VERBOSE=...`, `MAX_RETRIES=...`, `STOP_ON_FAILURE=...`, `MODEL=...`, `PRIZMKIT_EFFORT=...`) before `python3`.
70
+ - Background daemon: place environment assignments inside one `--env "..."` value.
71
+ - **Parameterized `MODEL` quoting**:
72
+ - Foreground: single-quote parameterized `MODEL` values containing colons, square brackets, parentheses, spaces, commas, or shell metacharacters:
73
+ ```bash
74
+ MODEL='<model-with-level>' VERBOSE=1 MAX_RETRIES=3 \
75
+ python3 ./.prizmkit/dev-pipeline/cli.py run bugfix .prizmkit/plans/bug-fix-list.json
76
+ ```
77
+ - Background daemon: wrap `--env` in double quotes and single-quote the `MODEL` value inside it to preserve non-space parameterized syntax (colons, square brackets, parentheses, commas) on both POSIX and Windows:
78
+ ```bash
79
+ python3 ./.prizmkit/dev-pipeline/cli.py daemon bugfix start .prizmkit/plans/bug-fix-list.json \
80
+ --env "VERBOSE=1 MAX_RETRIES=3 MODEL='<model-with-level>'"
81
+ ```
82
+ If a `MODEL` value contains spaces, quote the whole `MODEL=...` assignment inside `--env` (for example, `--env "VERBOSE=1 MAX_RETRIES=3 'MODEL=<model with spaces>'"`) so both POSIX and Windows daemon env parsing preserve spaces cleanly.
42
83
  - Include only confirmed options.
43
84
  - Preserve the user-provided list path.
44
85
  - Shell-quote custom paths and values.
@@ -1,20 +1,23 @@
1
1
  ---
2
2
  name: "feature-pipeline-launcher"
3
- description: "Build feature-pipeline commands from a validated .prizmkit/plans/feature-list.json without executing them. Produces foreground or background-daemon start commands plus status, logs, stop, and safe retry commands. Use whenever the user wants to run the feature pipeline, start building from an existing feature list, construct a feature subset command, inspect pipeline status, view logs, stop a run, or retry a feature. Trigger on: 'run pipeline', 'start pipeline', 'start building', 'build feature list', 'pipeline status', 'pipeline logs', 'stop pipeline', 'retry feature', 'clean retry F-001'."
3
+ description: "Build feature-pipeline commands from a validated .prizmkit/plans/feature-list.json without executing the pipeline. Supports bounded read-only AI CLI model/effort capability probing on demand and produces foreground or background-daemon start commands plus status, logs, stop, and safe retry commands. Use whenever the user wants to run the feature pipeline, start building from an existing feature list, construct a feature subset command, inspect pipeline status, view logs, stop a run, or retry a feature. Trigger on: 'run pipeline', 'start pipeline', 'start building', 'build feature list', 'pipeline status', 'pipeline logs', 'stop pipeline', 'retry feature', 'clean retry F-001'."
4
4
  ---
5
5
 
6
6
  # Feature Pipeline Command Builder
7
7
 
8
- Build commands for the autonomous feature pipeline from within an AI CLI conversation. This skill reads configuration and planning inputs, asks only the choices needed to construct an accurate command, displays the command and its effects, then stops.
8
+ Build commands for the autonomous feature pipeline from within an AI CLI conversation. This skill reads configuration and planning inputs, runs bounded read-only AI CLI capability probes only when custom model or reasoning effort configuration is needed, asks only the choices required to construct an accurate command, displays the command and its effects, then stops.
9
9
 
10
- ## Command-Only Boundary
10
+ ## Execution Boundary
11
11
 
12
- This skill:
12
+ This skill separates bounded read-only configuration probing from pipeline execution:
13
13
 
14
- - reads a validated feature list and relevant configuration;
15
- - builds prerequisite-check, start, status, logs, stop, and retry commands;
16
- - explains command effects and destructive options before returning them;
17
- - never executes shell commands, starts processes, follows logs, changes pipeline state, implements source code, or runs verification.
14
+ - **Allowed (bounded read-only configuration probing only)**:
15
+ - read a validated feature list and relevant configuration;
16
+ - build prerequisite-check, start, status, logs, stop, and retry commands;
17
+ - explain command effects and destructive options before returning them;
18
+ - only when the user selects Advanced config or explicitly mentions/asks about a custom model or reasoning effort: run bounded, side-effect-free read-only probes — `python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli`, check whether the resolved CLI executable exists on `PATH`, and inspect `<cli> --help` (plus `<cli> exec --help` when applicable). Read-only probes must never invoke prompt runs, interactive logins, file writes, Git mutations, or pipeline state changes.
19
+ - **Prohibited (pipeline execution and state mutation)**:
20
+ - never execute pipeline commands (`run`, `daemon start`, `stop`, `reset`), start processes, follow logs, change pipeline or Git state, implement source code, or run verification unless explicitly requested by the user.
18
21
 
19
22
  Do not invoke another workflow or implementation skill from this command builder. The user decides where and when to run the returned command.
20
23
 
@@ -25,7 +28,7 @@ Use this skill for:
25
28
  - constructing a foreground or background-daemon feature pipeline command;
26
29
  - selecting a feature subset such as `F-001:F-005` or `F-001,F-003`;
27
30
  - constructing status, logs, stop, or retry commands;
28
- - translating user-selected runtime options into environment variables.
31
+ - translating user-selected runtime options (`MODEL`, adaptive `PRIZMKIT_EFFORT`, retries, logging) into environment variables.
29
32
 
30
33
  Do not use it to:
31
34
 
@@ -82,7 +85,9 @@ Use a separate `AskUserQuestion` call with exactly three questions:
82
85
  2. **Max retries** — 3, 1, 5, or a user-provided non-negative integer.
83
86
  3. **Advanced config** — No or Yes.
84
87
 
85
- If Advanced config is Yes, load the advanced question set from `${SKILL_DIR}/references/configuration.md`. Advanced configuration contains only stop-on-failure and reasoning effort.
88
+ If Advanced config is No and the user did not specify or ask about a custom model or reasoning effort, skip read-only CLI probing (`diagnostics ai-cli` and `--help`) and record `Read-only CLI probe: skipped (default launch without custom model/effort)`.
89
+
90
+ If Advanced config is Yes — or if the user specified or asked about a custom model or reasoning effort — load `${SKILL_DIR}/references/configuration.md`, run its on-demand read-only CLI capability probe first, and collect confirmed advanced options (stop-on-failure, `MODEL`, and adaptive `PRIZMKIT_EFFORT`) following its four-branch placement rules.
86
91
 
87
92
  ### Step 4: Build command fragments
88
93
 
@@ -94,7 +99,7 @@ Build the prerequisite-check command separately from the pipeline command so the
94
99
  python3 ./.prizmkit/dev-pipeline/cli.py diagnostics version && git --version
95
100
  ```
96
101
 
97
- Also include the configured AI CLI check fragment from the reference. These are returned commands, not checks performed by this skill.
102
+ Also include the configured AI CLI check fragment from the reference so the user can verify prerequisites in their own shell. Even when on-demand read-only probing ran during configuration, pipeline commands are never executed by this skill.
98
103
 
99
104
  ### Step 5: Return the start command
100
105
 
@@ -126,7 +131,7 @@ python3 ./.prizmkit/dev-pipeline/cli.py daemon feature start .prizmkit/plans/fea
126
131
  --features F-001:F-005 --env "VERBOSE=1 MAX_RETRIES=3"
127
132
  ```
128
133
 
129
- Return only the template matching the confirmed mode and options. Explain that the command has not been executed, then stop.
134
+ Return only the template matching the confirmed mode and options. Explain that pipeline execution was not performed by this skill, then stop.
130
135
 
131
136
  ## Status Command
132
137
 
@@ -217,6 +222,7 @@ For every request, return:
217
222
  2. Input plan path and selected IDs, when applicable.
218
223
  3. Assumptions and safety warnings.
219
224
  4. Exact command block.
220
- 5. `Command execution: not performed by this skill.`
225
+ 5. `Read-only CLI probe: <diagnostics-and-help-summary | skipped (default launch without custom model/effort)>`
226
+ 6. `Pipeline execution: not performed by this skill; run any command above manually when ready.`
221
227
 
222
- Never report command success, process IDs, log locations, task completion, verification results, or state changes unless the user later supplies authoritative runtime output for explanation.
228
+ Never report pipeline execution success, process IDs, log locations, task completion, verification results, or state changes unless the user later supplies authoritative runtime output for explanation.
@@ -1,10 +1,10 @@
1
1
  # Configuration Reference — Feature Pipeline Command Builder
2
2
 
3
- Use this reference only to translate confirmed user choices into command fragments. The skill returns commands; it does not execute them.
3
+ Use this reference to run bounded on-demand read-only AI CLI capability probing and translate confirmed user choices into command fragments. This skill constructs and returns pipeline commands; it never executes pipeline runs or mutates state unless explicitly requested by the user.
4
4
 
5
5
  ## Configured AI CLI Prerequisite Command
6
6
 
7
- Runtime AI CLI selection is resolved by the canonical runtime across `AI_CLI`, `.prizmkit/config.json` (`ai_cli`), and installed host configuration in the exact order used by `cli.py run`.
7
+ Runtime AI CLI selection is resolved by the canonical runtime across `AI_CLI`, `.prizmkit/config.json` (`ai_cli` and `ai_cli_launch`), `.prizmkit/manifest.json` (`platform`), and installed host configuration in the exact order used by `cli.py run`.
8
8
 
9
9
  Return this cross-platform check command as part of start-command preparation:
10
10
 
@@ -12,7 +12,16 @@ Return this cross-platform check command as part of start-command preparation:
12
12
  python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
13
13
  ```
14
14
 
15
- `diagnostics ai-cli` reports the resolved CLI, source, and runtime mode, and exits non-zero if no configured AI CLI executable is available. Do not claim the configured executable exists until the user supplies successful command output.
15
+ `diagnostics ai-cli` reports `detected_cli`, `platform`, `source`, `runtime_mode`, `model_arg_supported`, `effort_flag_mode`, `supported_effort_values`, and `cli_available`, and exits non-zero if no AI CLI can be resolved (`detected_cli: (not found)`).
16
+
17
+ ## On-Demand Read-Only CLI Capability Probe
18
+
19
+ - **Fast default launch (skip probing)**: When Advanced config is not selected and the user did not mention or ask about a custom model or reasoning effort, skip all read-only CLI probing (`diagnostics ai-cli` and `--help`) and output `Read-only CLI probe: skipped (default launch without custom model/effort)`.
20
+ - **On-demand probe trigger**: Run the bounded read-only probe sequence only when Advanced config is selected or the user explicitly specifies or asks about `MODEL` or `PRIZMKIT_EFFORT` (including when paired with a per-invocation `AI_CLI=<cli>` override):
21
+ 1. Run `python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli` — or `AI_CLI=<cli> python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli` whenever the user explicitly specifies a per-invocation `AI_CLI=<cli>` override (and preserve `AI_CLI=<cli>` in the returned launch command) — then read the structured capability fields directly: `detected_cli`, `platform`, `source`, `model_arg_supported` (`true`|`false`), `effort_flag_mode` (`flag`|`config`|`none`|`custom_ignored`), `supported_effort_values` (comma-separated values or `none`), and `cli_available` (`true`|`false`). Consume these structured fields directly with zero legacy fallback parsing; if any required capability field is absent, treat the probe as failed under Branch 1.
22
+ 2. If `effort_flag_mode` is `custom_ignored`, route directly to Branch 2 without requiring `<cli> --help`.
23
+ 3. Otherwise, verify that `cli_available` is `true` and the resolved `detected_cli` executable exists on `PATH` (when not `(not found)`), then run `<cli> --help` (and `<cli> exec --help` when the CLI exposes an `exec` subcommand or `effort_flag_mode` is `config`) to inspect live model flags, independent effort/thinking options, or model-embedded intensity syntax.
24
+ 4. Keep every probe strictly read-only and bounded: never invoke prompt runs, interactive logins, file writes, Git mutations, or pipeline commands (`run`, `daemon start`, `stop`, `reset`). Record a concise summary of observed capability facts in `Read-only CLI probe:`.
16
25
 
17
26
  ## Base Configuration Mapping
18
27
 
@@ -24,21 +33,53 @@ python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
24
33
 
25
34
  ## Advanced Configuration Questions
26
35
 
27
- Ask only when the user selected Advanced config.
36
+ Ask only when the user selected Advanced config or explicitly requested custom model/effort configuration. Configure `MODEL` before `PRIZMKIT_EFFORT`:
28
37
 
29
38
  1. **Stop on failure**
30
39
  - Off (default) — continue to other eligible tasks after a task exhausts retries.
31
40
  - On — add `STOP_ON_FAILURE=1`.
32
- 2. **Reasoning effort**
41
+ 2. **Model (`MODEL`)**
42
+ - Default — omit `MODEL` so the resolved AI CLI uses its default model.
43
+ - Explicit model — add `MODEL=<name>` (or single-quoted `MODEL='<model-with-level>'` for parameterized model syntax) when `model_arg_supported` is `true` (or when probing failed without diagnostics confirming `model_arg_supported: false` and the user explicitly confirms a model name).
44
+ - If `diagnostics ai-cli` reports `effort_flag_mode: custom_ignored`, warn clearly that the active custom launch profile ignores both `MODEL` and `PRIZMKIT_EFFORT` (pass provider flags via `.prizmkit/config.json` `ai_cli_launch.extra_args`). If `model_arg_supported: false` is reported for an unresolved/unsupported profile, warn that `MODEL` is not forwarded and leave it unset.
45
+ - **Per-item precedence**: an individual feature's `"model"` field in `.prizmkit/plans/feature-list.json` takes precedence over the global `MODEL` environment variable during task execution.
46
+ 3. **Reasoning effort (`PRIZMKIT_EFFORT`)**
33
47
  - Default — omit `PRIZMKIT_EFFORT`.
34
- - `low`, `medium`, `high`, `xhigh`, or `max` — add `PRIZMKIT_EFFORT=<value>` after confirming the configured AI CLI supports it.
48
+ - Adaptive placement — determine whether and how to emit `PRIZMKIT_EFFORT` strictly from the four ordered placement branches below (never hardcode one universal effort list or CLI allowlist).
49
+
50
+ ## Adaptive Placement Branches (Evaluate 1 → 4 Strictly in Order)
51
+
52
+ Evaluate branches strictly from 1 to 4 using `diagnostics ai-cli` capability fields plus live `--help` output; the first matching branch governs placement. Never branch on hardcoded platform CLI allowlists.
35
53
 
36
- These are the complete advanced choices for this command builder.
54
+ 1. **Branch 1 — Probe failed, CLI missing, or insufficient help output**:
55
+ - Condition: `diagnostics ai-cli` fails or exits non-zero, `detected_cli` is `(not found)` or missing on `PATH`, `cli_available` is `false`, required capability fields (`model_arg_supported`, `effort_flag_mode`, `supported_effort_values`, `cli_available`) are absent, OR (when `effort_flag_mode` is not `custom_ignored`) `<cli> --help` fails or yields insufficient output.
56
+ - Action: Never guess or synthesize `PRIZMKIT_EFFORT`. Omit `PRIZMKIT_EFFORT`, state the exact probe gap in `Read-only CLI probe:` and warnings, and allow an optional explicit `MODEL=<name>` only if diagnostics did not confirm `model_arg_supported: false` (otherwise leave `MODEL` unset).
57
+ 2. **Branch 2 — Model-embedded intensity syntax or `custom_ignored` profile**:
58
+ - Condition: `effort_flag_mode` is `custom_ignored` (routed directly without letting `<cli> --help` failure trigger Branch 1), OR (`model_arg_supported` is `true`, no compatible independent effort mechanism applies — `effort_flag_mode` is `none`, or `supported_effort_values` is `none`, or `effort_flag_mode` is `flag` with no compatible effort/thinking flag in `--help`, excluding compatible `effort_flag_mode: config` — AND live `--help` documents encoding intensity inside the model identifier syntax such as `<model>:<level>` or parameterized model notation).
59
+ - Action: Omit `PRIZMKIT_EFFORT`. Under `custom_ignored`, warn clearly that the active custom launch profile ignores both `MODEL` and `PRIZMKIT_EFFORT` (configure provider flags via `.prizmkit/config.json` `ai_cli_launch.extra_args`), never emit `PRIZMKIT_EFFORT` (the runtime rejects non-empty `PRIZMKIT_EFFORT` pre-launch on custom profiles), and omit `MODEL` unless the user explicitly overrides. Under model-embedded intensity syntax, guide the user to set `MODEL=<model-with-level>` (or leave `MODEL` unset for the CLI default).
60
+ 3. **Branch 3 — Independent effort explicitly unsupported**:
61
+ - Condition: Branches 1 and 2 do not match, AND `effort_flag_mode` is `none`, OR `supported_effort_values` is `none`, OR `effort_flag_mode` is `flag` yet live `--help` shows no compatible effort/thinking option (excluding compatible `effort_flag_mode: config`).
62
+ - Action: Do not emit `PRIZMKIT_EFFORT` (keep it unset so the runtime does not reject it pre-launch). Still permit `MODEL=<name>` when `model_arg_supported` is `true`.
63
+ 4. **Branch 4 — Independent effort compatible**:
64
+ - Condition: (`effort_flag_mode` is `flag` with live `--help` confirming a compatible effort/thinking option, OR `effort_flag_mode` is `config`) AND `supported_effort_values` is non-empty (not `none`).
65
+ - Action: When live `--help` enumerates concrete effort/thinking values, intersect those enumerated choices with `supported_effort_values`; if the intersection is empty, omit `PRIZMKIT_EFFORT` and disclose the capability mismatch rather than emitting an unvalidated value. Otherwise present the valid choices and emit `PRIZMKIT_EFFORT=<value>` (or omit for default). If the confirmed `MODEL` value already embeds a reasoning-effort suffix or parameter (for CLIs that accept both model-embedded intensity and an independent effort flag), omit `PRIZMKIT_EFFORT` by default and note that intensity is already specified in `MODEL` so conflicting intensity settings are not sent together.
37
66
 
38
67
  ## Command Construction Rules
39
68
 
40
- - Foreground: place environment assignments before `python3`.
41
- - Background daemon: place assignments inside one `--env "..."` value.
69
+ - Foreground: place environment assignments (`AI_CLI=...` when overridden, `VERBOSE=...`, `MAX_RETRIES=...`, `STOP_ON_FAILURE=...`, `MODEL=...`, `PRIZMKIT_EFFORT=...`) before `python3`.
70
+ - Background daemon: place environment assignments inside one `--env "..."` value.
71
+ - **Parameterized `MODEL` quoting**:
72
+ - Foreground: single-quote parameterized `MODEL` values containing colons, square brackets, parentheses, spaces, commas, or shell metacharacters:
73
+ ```bash
74
+ MODEL='<model-with-level>' VERBOSE=1 MAX_RETRIES=3 \
75
+ python3 ./.prizmkit/dev-pipeline/cli.py run feature .prizmkit/plans/feature-list.json
76
+ ```
77
+ - Background daemon: wrap `--env` in double quotes and single-quote the `MODEL` value inside it to preserve non-space parameterized syntax (colons, square brackets, parentheses, commas) on both POSIX and Windows:
78
+ ```bash
79
+ python3 ./.prizmkit/dev-pipeline/cli.py daemon feature start .prizmkit/plans/feature-list.json \
80
+ --env "VERBOSE=1 MAX_RETRIES=3 MODEL='<model-with-level>'"
81
+ ```
82
+ If a `MODEL` value contains spaces, quote the whole `MODEL=...` assignment inside `--env` (for example, `--env "VERBOSE=1 MAX_RETRIES=3 'MODEL=<model with spaces>'"`) so both POSIX and Windows daemon env parsing preserve spaces cleanly.
42
83
  - Include only confirmed options.
43
84
  - Preserve the user-provided list path and feature filter.
44
85
  - Shell-quote custom paths and values.
@@ -1,20 +1,23 @@
1
1
  ---
2
2
  name: "iterate-pipeline-launcher"
3
- description: "Build autonomous product-iteration commands without executing them. Produces foreground or background-daemon start commands plus status, logs, stop, and post-pause resume commands in both short prizmkit and canonical Python Runtime forms. Use whenever the user wants to start autonomous product iteration, set a focus direction or round limit, inspect iterate status, view iterate logs, stop an iterate daemon, or resume after a checkpoint or user-confirmation pause. Trigger on: 'iterate product', 'autonomous iteration', 'run iterate', 'start iterate', 'iterate status', 'iterate logs', 'stop iterate', 'resume iterate', '自主迭代', '产品迭代'."
3
+ description: "Build autonomous product-iteration commands without executing the pipeline. Supports bounded read-only AI CLI model/effort capability probing on demand and produces foreground or background-daemon start commands plus status, logs, stop, and post-pause resume commands in both short prizmkit and canonical Python Runtime forms. Use whenever the user wants to start autonomous product iteration, set a focus direction or round limit, inspect iterate status, view iterate logs, stop an iterate daemon, or resume after a checkpoint or user-confirmation pause. Trigger on: 'iterate product', 'autonomous iteration', 'run iterate', 'start iterate', 'iterate status', 'iterate logs', 'stop iterate', 'resume iterate', '自主迭代', '产品迭代'."
4
4
  ---
5
5
 
6
6
  # Iterate Pipeline Command Builder
7
7
 
8
- Build commands for the autonomous product-iteration loop from within an AI CLI conversation. This skill inspects optional project and iterate state for context, asks only the choices needed to construct accurate commands, displays both the short `prizmkit` form and the canonical Python Runtime form with their effects, then stops.
8
+ Build commands for the autonomous product-iteration loop from within an AI CLI conversation. This skill inspects optional project and iterate state for context, runs bounded read-only AI CLI capability probes only when custom model or reasoning effort configuration is needed, asks only the choices required to construct accurate commands, displays both the short `prizmkit` form and the canonical Python Runtime form with their effects, then stops.
9
9
 
10
- ## Command-Only Boundary
10
+ ## Execution Boundary
11
11
 
12
- This skill:
12
+ This skill separates bounded read-only configuration probing from pipeline execution:
13
13
 
14
- - inspects optional project context (`.prizmkit/plans/feature-list.json`, `.prizmkit/state/iterate/state.json`, `.prizmkit/state/iterate/pending-decisions.md`) when present;
15
- - builds prerequisite-check, start, status, logs, stop, and post-pause resume commands;
16
- - explains safety caps, checkpoint pauses, and command effects before returning them;
17
- - never executes shell commands, starts processes, follows logs, changes pipeline state, implements source code, or runs verification.
14
+ - **Allowed (bounded read-only configuration probing only)**:
15
+ - inspect optional project context (`.prizmkit/plans/feature-list.json`, `.prizmkit/state/iterate/state.json`, `.prizmkit/state/iterate/pending-decisions.md`) when present;
16
+ - build prerequisite-check, start, status, logs, stop, and post-pause resume commands;
17
+ - explain safety caps, checkpoint pauses, and command effects before returning them;
18
+ - only when the user selects Advanced config or explicitly mentions/asks about a custom model or reasoning effort: run bounded, side-effect-free read-only probes — `python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli`, check whether the resolved CLI executable exists on `PATH`, and inspect `<cli> --help` (plus `<cli> exec --help` when applicable). Read-only probes must never invoke prompt runs, interactive logins, file writes, Git mutations, or pipeline state changes.
19
+ - **Prohibited (pipeline execution and state mutation)**:
20
+ - never execute pipeline commands (`run`, `daemon start`, `stop`, `reset`), start processes, follow logs, change pipeline or Git state, implement source code, or run verification unless explicitly requested by the user.
18
21
 
19
22
  Do not invoke another workflow or implementation skill from this command builder. The user decides where and when to run the returned command.
20
23
 
@@ -23,7 +26,7 @@ Do not invoke another workflow or implementation skill from this command builder
23
26
  Use this skill for:
24
27
 
25
28
  - constructing a foreground or background-daemon autonomous product-iteration start command;
26
- - configuring optional `--focus`, round limit (`--max-rounds` or AI autonomous judgment), verbose logging, and advanced checkpoint or retry options;
29
+ - configuring optional `--focus`, round limit (`--max-rounds` or AI autonomous judgment), verbose logging, `MODEL`, adaptive `PRIZMKIT_EFFORT`, and advanced checkpoint or retry options;
27
30
  - constructing iterate status, logs, stop, or post-pause resume commands;
28
31
  - translating user-selected runtime options into flags and environment variables.
29
32
 
@@ -84,7 +87,9 @@ Use a separate `AskUserQuestion` call to collect the user's iteration preference
84
87
  3. **Verbose logging** — Off or On.
85
88
  4. **Advanced config** — No (default) or Yes.
86
89
 
87
- If Advanced config is Yes, load the advanced question set from `${SKILL_DIR}/references/configuration.md` (`--checkpoint-every`, `--no-checkpoint`, `MAX_RETRIES`, `PRIZMKIT_EFFORT`, and optional `PRIZMKIT_ITERATE_SAFETY_CAP`).
90
+ If Advanced config is No and the user did not specify or ask about a custom model or reasoning effort, skip read-only CLI probing (`diagnostics ai-cli` and `--help`) and record `Read-only CLI probe: skipped (default launch without custom model/effort)`.
91
+
92
+ If Advanced config is Yes — or if the user specified or asked about a custom model or reasoning effort — load `${SKILL_DIR}/references/configuration.md`, run its on-demand read-only CLI capability probe first, and collect confirmed advanced options (`--checkpoint-every`, `--no-checkpoint`, `MAX_RETRIES`, `MODEL`, adaptive `PRIZMKIT_EFFORT`, and optional `PRIZMKIT_ITERATE_SAFETY_CAP`) following its four-branch placement rules.
88
93
 
89
94
  ### Step 4: Build command fragments
90
95
 
@@ -106,7 +111,7 @@ python3 ./.prizmkit/dev-pipeline/cli.py diagnostics version && git --version
106
111
  python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
107
112
  ```
108
113
 
109
- Always include the configured AI CLI check fragment from the reference. These are returned commands, not checks performed by this skill.
114
+ Always include the configured AI CLI check fragment from the reference so the user can verify prerequisites in their own shell. Even when on-demand read-only probing ran during configuration, pipeline commands are never executed by this skill.
110
115
 
111
116
  ### Step 5: Return the start command
112
117
 
@@ -163,7 +168,7 @@ python3 ./.prizmkit/dev-pipeline/cli.py daemon iterate start --focus "Improve on
163
168
  --env "VERBOSE=1 MAX_RETRIES=3"
164
169
  ```
165
170
 
166
- Explain that the command has not been executed, then stop.
171
+ Explain that pipeline execution was not performed by this skill, then stop.
167
172
 
168
173
  ## Status Command
169
174
 
@@ -262,6 +267,7 @@ For every request, return:
262
267
  2. Relevant optional context summary (existing backlog or prior pause state, if observed).
263
268
  3. Assumptions, safety-cap/checkpoint disclosures, and warnings.
264
269
  4. Exact command block showing both short `prizmkit` and canonical `python3 ./.prizmkit/dev-pipeline/cli.py` forms.
265
- 5. `Command execution: not performed by this skill.`
270
+ 5. `Read-only CLI probe: <diagnostics-and-help-summary | skipped (default launch without custom model/effort)>`
271
+ 6. `Pipeline execution: not performed by this skill; run any command above manually when ready.`
266
272
 
267
- Never report command success, process IDs, log locations, round completion, verification results, or state changes unless the user later supplies authoritative runtime output for explanation.
273
+ Never report pipeline execution success, process IDs, log locations, round completion, verification results, or state changes unless the user later supplies authoritative runtime output for explanation.
@@ -1,10 +1,10 @@
1
1
  # Configuration Reference — Iterate Pipeline Command Builder
2
2
 
3
- Use this reference only to translate confirmed user choices into command fragments. The skill returns commands; it does not execute them.
3
+ Use this reference to run bounded on-demand read-only AI CLI capability probing and translate confirmed user choices into command fragments. This skill constructs and returns pipeline commands; it never executes pipeline runs or mutates state unless explicitly requested by the user.
4
4
 
5
5
  ## Configured AI CLI Prerequisite Command
6
6
 
7
- Runtime AI CLI selection is resolved by the canonical runtime across `AI_CLI`, `.prizmkit/config.json` (`ai_cli`), and installed host configuration in the exact order used by `cli.py run`.
7
+ Runtime AI CLI selection is resolved by the canonical runtime across `AI_CLI`, `.prizmkit/config.json` (`ai_cli` and `ai_cli_launch`), `.prizmkit/manifest.json` (`platform`), and installed host configuration in the exact order used by `cli.py run`.
8
8
 
9
9
  Return this cross-platform check command as part of start-command preparation (short and canonical forms):
10
10
 
@@ -13,7 +13,16 @@ prizmkit diagnostics ai-cli
13
13
  python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
14
14
  ```
15
15
 
16
- `diagnostics ai-cli` reports the resolved CLI, source, and runtime mode, and exits non-zero if no configured AI CLI executable is available. Do not claim the configured executable exists until the user supplies successful command output.
16
+ `diagnostics ai-cli` reports `detected_cli`, `platform`, `source`, `runtime_mode`, `model_arg_supported`, `effort_flag_mode`, `supported_effort_values`, and `cli_available`, and exits non-zero if no AI CLI can be resolved (`detected_cli: (not found)`).
17
+
18
+ ## On-Demand Read-Only CLI Capability Probe
19
+
20
+ - **Fast default launch (skip probing)**: When Advanced config is not selected and the user did not mention or ask about a custom model or reasoning effort, skip all read-only CLI probing (`diagnostics ai-cli` and `--help`) and output `Read-only CLI probe: skipped (default launch without custom model/effort)`.
21
+ - **On-demand probe trigger**: Run the bounded read-only probe sequence only when Advanced config is selected or the user explicitly specifies or asks about `MODEL` or `PRIZMKIT_EFFORT` (including when paired with a per-invocation `AI_CLI=<cli>` override):
22
+ 1. Run `python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli` — or `AI_CLI=<cli> python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli` whenever the user explicitly specifies a per-invocation `AI_CLI=<cli>` override (and preserve `AI_CLI=<cli>` in the returned launch command) — then read the structured capability fields directly: `detected_cli`, `platform`, `source`, `model_arg_supported` (`true`|`false`), `effort_flag_mode` (`flag`|`config`|`none`|`custom_ignored`), `supported_effort_values` (comma-separated values or `none`), and `cli_available` (`true`|`false`). Consume these structured fields directly with zero legacy fallback parsing; if any required capability field is absent, treat the probe as failed under Branch 1.
23
+ 2. If `effort_flag_mode` is `custom_ignored`, route directly to Branch 2 without requiring `<cli> --help`.
24
+ 3. Otherwise, verify that `cli_available` is `true` and the resolved `detected_cli` executable exists on `PATH` (when not `(not found)`), then run `<cli> --help` (and `<cli> exec --help` when the CLI exposes an `exec` subcommand or `effort_flag_mode` is `config`) to inspect live model flags, independent effort/thinking options, or model-embedded intensity syntax.
25
+ 4. Keep every probe strictly read-only and bounded: never invoke prompt runs, interactive logins, file writes, Git mutations, or pipeline commands (`run`, `daemon start`, `stop`, `reset`). Record a concise summary of observed capability facts in `Read-only CLI probe:`.
17
26
 
18
27
  ## Base Configuration Mapping
19
28
 
@@ -28,7 +37,7 @@ python3 ./.prizmkit/dev-pipeline/cli.py diagnostics ai-cli
28
37
 
29
38
  ## Advanced Configuration Questions
30
39
 
31
- Ask only when the user selected Advanced config.
40
+ Ask only when the user selected Advanced config or explicitly requested custom model/effort configuration. Configure `MODEL` before `PRIZMKIT_EFFORT`:
32
41
 
33
42
  1. **Checkpoint pause behavior**
34
43
  - Default — pause every 5 completed rounds in autonomous mode (omitted when `--max-rounds` is explicitly set).
@@ -37,19 +46,61 @@ Ask only when the user selected Advanced config.
37
46
  2. **Max retries per feature**
38
47
  - Default (3) — omit `MAX_RETRIES` or set `MAX_RETRIES=3`.
39
48
  - Custom non-negative integer N — add `MAX_RETRIES=N`.
40
- 3. **Reasoning effort**
49
+ 3. **Model (`MODEL`)**
50
+ - Default — omit `MODEL` so the resolved AI CLI uses its default model.
51
+ - Explicit model — add `MODEL=<name>` (or single-quoted `MODEL='<model-with-level>'` for parameterized model syntax) when `model_arg_supported` is `true` (or when probing failed without diagnostics confirming `model_arg_supported: false` and the user explicitly confirms a model name).
52
+ - If `diagnostics ai-cli` reports `effort_flag_mode: custom_ignored`, warn clearly that the active custom launch profile ignores both `MODEL` and `PRIZMKIT_EFFORT` (pass provider flags via `.prizmkit/config.json` `ai_cli_launch.extra_args`). If `model_arg_supported: false` is reported for an unresolved/unsupported profile, warn that `MODEL` is not forwarded and leave it unset.
53
+ - **Per-item precedence**: global `MODEL` governs Phase A/B iterate evaluation and headless planning sessions and serves as the default for feature execution, while any individual feature item's `"model"` field in `.prizmkit/plans/feature-list.json` takes precedence over global `MODEL` when that feature runs.
54
+ 4. **Reasoning effort (`PRIZMKIT_EFFORT`)**
41
55
  - Default — omit `PRIZMKIT_EFFORT`.
42
- - `low`, `medium`, `high`, `xhigh`, or `max` — add `PRIZMKIT_EFFORT=<value>` after confirming the configured AI CLI supports it.
43
- 4. **Autonomous safety cap (optional)**
56
+ - Adaptive placement — determine whether and how to emit `PRIZMKIT_EFFORT` strictly from the four ordered placement branches below (never hardcode one universal effort list or CLI allowlist).
57
+ 5. **Autonomous safety cap (optional)**
44
58
  - Default (10 rounds) — omit `PRIZMKIT_ITERATE_SAFETY_CAP`.
45
59
  - Custom positive integer N — add `PRIZMKIT_ITERATE_SAFETY_CAP=N` (applies only in autonomous mode when `--max-rounds` is omitted; coinciding thresholds prioritize `PAUSE_SAFETY_CAP` over `PAUSE_CHECKPOINT_REVIEW`).
46
60
 
47
- These are the complete advanced choices for this command builder.
61
+ ## Adaptive Placement Branches (Evaluate 1 → 4 Strictly in Order)
62
+
63
+ Evaluate branches strictly from 1 to 4 using `diagnostics ai-cli` capability fields plus live `--help` output; the first matching branch governs placement. Never branch on hardcoded platform CLI allowlists.
64
+
65
+ 1. **Branch 1 — Probe failed, CLI missing, or insufficient help output**:
66
+ - Condition: `diagnostics ai-cli` fails or exits non-zero, `detected_cli` is `(not found)` or missing on `PATH`, `cli_available` is `false`, required capability fields (`model_arg_supported`, `effort_flag_mode`, `supported_effort_values`, `cli_available`) are absent, OR (when `effort_flag_mode` is not `custom_ignored`) `<cli> --help` fails or yields insufficient output.
67
+ - Action: Never guess or synthesize `PRIZMKIT_EFFORT`. Omit `PRIZMKIT_EFFORT`, state the exact probe gap in `Read-only CLI probe:` and warnings, and allow an optional explicit `MODEL=<name>` only if diagnostics did not confirm `model_arg_supported: false` (otherwise leave `MODEL` unset).
68
+ 2. **Branch 2 — Model-embedded intensity syntax or `custom_ignored` profile**:
69
+ - Condition: `effort_flag_mode` is `custom_ignored` (routed directly without letting `<cli> --help` failure trigger Branch 1), OR (`model_arg_supported` is `true`, no compatible independent effort mechanism applies — `effort_flag_mode` is `none`, or `supported_effort_values` is `none`, or `effort_flag_mode` is `flag` with no compatible effort/thinking flag in `--help`, excluding compatible `effort_flag_mode: config` — AND live `--help` documents encoding intensity inside the model identifier syntax such as `<model>:<level>` or parameterized model notation).
70
+ - Action: Omit `PRIZMKIT_EFFORT`. Under `custom_ignored`, warn clearly that the active custom launch profile ignores both `MODEL` and `PRIZMKIT_EFFORT` (configure provider flags via `.prizmkit/config.json` `ai_cli_launch.extra_args`), never emit `PRIZMKIT_EFFORT` (the runtime rejects non-empty `PRIZMKIT_EFFORT` pre-launch on custom profiles), and omit `MODEL` unless the user explicitly overrides. Under model-embedded intensity syntax, guide the user to set `MODEL=<model-with-level>` (or leave `MODEL` unset for the CLI default).
71
+ 3. **Branch 3 — Independent effort explicitly unsupported**:
72
+ - Condition: Branches 1 and 2 do not match, AND `effort_flag_mode` is `none`, OR `supported_effort_values` is `none`, OR `effort_flag_mode` is `flag` yet live `--help` shows no compatible effort/thinking option (excluding compatible `effort_flag_mode: config`).
73
+ - Action: Do not emit `PRIZMKIT_EFFORT` (keep it unset so the runtime does not reject it pre-launch). Still permit `MODEL=<name>` when `model_arg_supported` is `true`.
74
+ 4. **Branch 4 — Independent effort compatible**:
75
+ - Condition: (`effort_flag_mode` is `flag` with live `--help` confirming a compatible effort/thinking option, OR `effort_flag_mode` is `config`) AND `supported_effort_values` is non-empty (not `none`).
76
+ - Action: When live `--help` enumerates concrete effort/thinking values, intersect those enumerated choices with `supported_effort_values`; if the intersection is empty, omit `PRIZMKIT_EFFORT` and disclose the capability mismatch rather than emitting an unvalidated value. Otherwise present the valid choices and emit `PRIZMKIT_EFFORT=<value>` (or omit for default). If the confirmed `MODEL` value already embeds a reasoning-effort suffix or parameter (for CLIs that accept both model-embedded intensity and an independent effort flag), omit `PRIZMKIT_EFFORT` by default and note that intensity is already specified in `MODEL` so conflicting intensity settings are not sent together.
48
77
 
49
78
  ## Command Construction Rules
50
79
 
51
- - Foreground (`run iterate`): place environment assignments (`VERBOSE=...`, `MAX_RETRIES=...`, `PRIZMKIT_EFFORT=...`, `PRIZMKIT_ITERATE_SAFETY_CAP=...`) before `prizmkit` or `python3`; omit env assignments when all defaults apply.
80
+ - Foreground (`run iterate`): place environment assignments (`AI_CLI=...` when overridden, `VERBOSE=...`, `MAX_RETRIES=...`, `MODEL=...`, `PRIZMKIT_EFFORT=...`, `PRIZMKIT_ITERATE_SAFETY_CAP=...`) before `prizmkit` or `python3`; omit env assignments when all defaults apply.
52
81
  - Background daemon (`daemon iterate start`): place environment assignments inside one `--env "..."` value; omit `--env` when no environment overrides are selected.
82
+ - **Parameterized `MODEL` quoting**:
83
+ - Foreground: single-quote parameterized `MODEL` values containing colons, square brackets, parentheses, spaces, commas, or shell metacharacters:
84
+ ```bash
85
+ # Short CLI
86
+ MODEL='<model-with-level>' VERBOSE=1 MAX_RETRIES=3 \
87
+ prizmkit run iterate
88
+
89
+ # Canonical Python Runtime
90
+ MODEL='<model-with-level>' VERBOSE=1 MAX_RETRIES=3 \
91
+ python3 ./.prizmkit/dev-pipeline/cli.py run iterate
92
+ ```
93
+ - Background daemon: wrap `--env` in double quotes and single-quote the `MODEL` value inside it to preserve non-space parameterized syntax (colons, square brackets, parentheses, commas) on both POSIX and Windows:
94
+ ```bash
95
+ # Short CLI
96
+ prizmkit daemon iterate start \
97
+ --env "VERBOSE=1 MAX_RETRIES=3 MODEL='<model-with-level>'"
98
+
99
+ # Canonical Python Runtime
100
+ python3 ./.prizmkit/dev-pipeline/cli.py daemon iterate start \
101
+ --env "VERBOSE=1 MAX_RETRIES=3 MODEL='<model-with-level>'"
102
+ ```
103
+ If a `MODEL` value contains spaces, quote the whole `MODEL=...` assignment inside `--env` (for example, `--env "VERBOSE=1 MAX_RETRIES=3 'MODEL=<model with spaces>'"`) so both POSIX and Windows daemon env parsing preserve spaces cleanly.
53
104
  - Do not pass `--dry-run` to `daemon iterate start`; `--dry-run` is supported only on foreground `run iterate`.
54
105
  - Include only confirmed options and shell-quote custom `--focus` values and paths.
55
106
  - Return only the confirmed execution mode (foreground or background daemon), providing both the short `prizmkit` command and the canonical `python3 ./.prizmkit/dev-pipeline/cli.py` command.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "prizmkit",
3
- "version": "1.1.194",
3
+ "version": "1.1.196",
4
4
  "description": "Create a new PrizmKit-powered project with clean initialization — no framework dev files, just what you need.",
5
5
  "type": "module",
6
6
  "bin": {