From 01d3c82131731e0f5f8d577c0682563f3ed9f1f7 Mon Sep 17 00:00:00 2001 From: Cursor Agent Date: Mon, 9 Nov 2026 03:00:54 +0000 Subject: [PATCH 1/2] Pass only the model id from a role line. Role lines stay grok-4.6 xhigh. spawn_subagent receives the model id, checked against grok models. This CLI has no effort argument, so the child keeps the parent's effort. Co-authored-by: Kaitian Xie --- skills/architect/SKILL.md | 2 +- skills/arena/SKILL.md | 2 +- skills/how/SKILL.md | 2 +- skills/poteto-mode/SKILL.md | 4 ++-- skills/reflect/SKILL.md | 2 +- skills/setup-pstack/SKILL.md | 4 ++-- skills/swarm/SKILL.md | 2 +- skills/why/SKILL.md | 2 +- 8 files changed, 10 insertions(+), 10 deletions(-) diff --git a/skills/architect/SKILL.md b/skills/architect/SKILL.md index 94d8682..d6fa3ee 100644 --- a/skills/architect/SKILL.md +++ b/skills/architect/SKILL.md @@ -28,7 +28,7 @@ Skip Phase A only when the work is genuinely greenfield with no surrounding syst Run the **arena** skill with the design-sketch task and the Phase A grounding artifacts. Pass `references/runner-prompt.md` as each runner's prompt. -Take the runners from the `architect runners` line in the `pstack-models.md` rule, in place of the `arena runners` line. If the rule or that line is missing, use `grok-4.6 xhigh` and `grok-4.5 high`. Alias and rejected entries follow the runner rules in the **arena** skill's Phase A. +Take the model id from each `architect runners` entry in the `pstack-models.md` rule, in place of the `arena runners` line, and validate it against `grok models`. If the rule or that line is missing, use `grok-4.6` at `xhigh` and `grok-4.5` at `high`. Alias and rejected entries follow the runner rules in the **arena** skill's Phase A. Design it twice. Require at least two structurally distinct candidates before synthesis, even when the first looks sufficient. This is the **exhaust-the-design-space** principle skill made concrete. Whole-shape alternatives, not point fixes inside one shape. diff --git a/skills/arena/SKILL.md b/skills/arena/SKILL.md index 17515c3..8127ee7 100644 --- a/skills/arena/SKILL.md +++ b/skills/arena/SKILL.md @@ -14,7 +14,7 @@ Every candidate gets the same prompt, so the prompt is the contract. 1. State the artifact each candidate is producing. 2. Derive the rubric, 3-6 concrete gradeable criteria for what success looks like on *this* task. It is the picker's tool in Phase D. Candidates only see the task. -3. Pick the runners. Use the `arena runners` line in `~/.grok/rules/pstack-models.md`. If the rule or that line is missing, default to one each on `grok-4.6 xhigh` and `grok-4.5 high`. An `auto` or `inherit-parent` entry in this line means the parent model, so omit `model` for it. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so, or drop the seat and say so when there is none. If dropped seats leave fewer than two runners, add runners until there are two, or one when the line has one entry. Run each added runner on a remaining runner's model, or omit `model` so it runs on the parent model when none remain, and say so. Diversity is a different detected slug, not a different vendor. Spawn more runners when the arena covers multiple design directions. Use the same model N times when the work is generation-bound rather than judgment-sensitive. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. +3. Pick the runners. Use the model id from each `arena runners` entry in `~/.grok/rules/pstack-models.md` (`grok-4.6` from `grok-4.6 xhigh`, `grok-4.5` from `grok-4.5 high`), validated against `grok models`. If the rule or that line is missing, default to one each on `grok-4.6` at `xhigh` and `grok-4.5` at `high`. An `auto` or `inherit-parent` entry in this line means the parent model, so omit `model` for it. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so, or drop the seat and say so when there is none. If dropped seats leave fewer than two runners, add runners until there are two, or one when the line has one entry. Run each added runner on a remaining runner's model, or omit `model` so it runs on the parent model when none remain, and say so. Diversity is a different detected slug, not a different vendor. Spawn more runners when the arena covers multiple design directions. Use the same model N times when the work is generation-bound rather than judgment-sensitive. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. 4. Give each candidate its own output location (a git worktree where possible, otherwise `/tmp/arena-/candidate-/`), per the **separate-before-serializing-shared-state** principle skill. ## Phase B: Fan out diff --git a/skills/how/SKILL.md b/skills/how/SKILL.md index d07dce5..7519c36 100644 --- a/skills/how/SKILL.md +++ b/skills/how/SKILL.md @@ -8,7 +8,7 @@ disable-model-invocation: true Explore the codebase to answer "how does X work?" questions. Produce architectural explanations at the level of a senior engineer onboarding onto a subsystem, enough to build a working mental model, not so much that it reads like annotated source code. -Every spawn is a `spawn_subagent` call with no `subagent_type` and a prompt that forbids file edits (there is no readonly flag). Explorers use the `how explorer` line in the `pstack-models.md` rule (default `grok-4.6` at `xhigh`), and every explainer uses the `how explainer` line (default `grok-4.6` at `xhigh`). Set `model` to that line's value, or to the default if the rule or the line is missing. Leave `model` unset when the value is `auto` or `inherit-parent`. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. +Every spawn is a `spawn_subagent` call with no `subagent_type` and a prompt that forbids file edits (there is no readonly flag). Explorers use the `how explorer` line in the `pstack-models.md` rule (default `grok-4.6` at `xhigh`), and every explainer uses the `how explainer` line (default `grok-4.6` at `xhigh`). Set `model` to the model id in that line's value (`grok-4.6` from `grok-4.6 xhigh`), or to the default model id if the rule or the line is missing. Validate the id against `grok models`. Leave `model` unset when the value is `auto` or `inherit-parent`. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. ## 1. Assess complexity diff --git a/skills/poteto-mode/SKILL.md b/skills/poteto-mode/SKILL.md index 5f6b944..63391a7 100644 --- a/skills/poteto-mode/SKILL.md +++ b/skills/poteto-mode/SKILL.md @@ -93,9 +93,9 @@ Read the leaf skill in full for any principle you apply. Each entry names when i **Use `subagent_type: "pstack:poteto-agent"` for any subagent you spawn inside a playbook step** (code-writing delegates, ad-hoc helpers). The chat that receives `/poteto-mode` runs the playbook and is the only chat that may call `spawn_subagent`. A child cannot spawn. Never spawn a `poteto-agent` to run the playbook itself. Routed workflow skills (`how`, `why`, `interrogate`, `reflect`, `swarm`) omit `subagent_type`, or set the type they name. Respect what the skill prescribes, don't override to `poteto-agent`. -**Defaults for every `spawn_subagent` call.** `run_in_background: true` (a steer or resume of a running agent too, because a foreground call holds your turn while other reports wait). There is no readonly flag. When a child must not edit files, say so in the prompt. MCP stays available through `search_tool` and `use_tool`. Pass file pointers, not inlined context. Pass `model` only when the tool schema includes it. Otherwise omit `model` so the child inherits the parent. If it rejects a slug, use the closest slug from `grok models` and say so. Defaults are `grok-4.6` at `xhigh` for code and for prose and judgment. Code delegates tier by difficulty. The hardest changes (cross-cutting design, gnarly concurrency, subtle algorithms) go to your strongest judgment model (`grok-4.6` at `xhigh`), whether the task needs judgment on vague intent or is a precisely specified sequence of steps to execute to the letter. Trivial mechanical edits go to your fast code model. Per-role lines in `~/.grok/rules/pstack-models.md` (written by `/setup-pstack`) override these defaults and the model choices in the routed skills (`how`, `why`, `arena`, `swarm`, `architect`, `interrogate`, `reflect`). A role with no line keeps its default, and a role line of `inherit-parent` or `auto` runs that role on the parent chat model (omit `model`). Each code playbook's configured model comes from its line (`feature, refactoring`, `bug-fix`, `perf-issue`, or `hillclimb`), and the hardest changes read `hardest tasks`. Prose and judgment read `judgment and prose`. When `spawn_subagent` has no `model` argument, omit `model` for this session so every role inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. +**Defaults for every `spawn_subagent` call.** `run_in_background: true` (a steer or resume of a running agent too, because a foreground call holds your turn while other reports wait). There is no readonly flag. When a child must not edit files, say so in the prompt. MCP stays available through `search_tool` and `use_tool`. Pass file pointers, not inlined context. Pass `model` only when the tool schema includes it. Otherwise omit `model` so the child inherits the parent. If it rejects a slug, use the closest slug from `grok models` and say so. Defaults are `grok-4.6` at `xhigh` for code and for prose and judgment. Code delegates tier by difficulty. The hardest changes (cross-cutting design, gnarly concurrency, subtle algorithms) go to your strongest judgment model (`grok-4.6` at `xhigh`), whether the task needs judgment on vague intent or is a precisely specified sequence of steps to execute to the letter. Trivial mechanical edits go to your fast code model. Per-role lines in `~/.grok/rules/pstack-models.md` (written by `/setup-pstack`) override these defaults and the model choices in the routed skills (`how`, `why`, `arena`, `swarm`, `architect`, `interrogate`, `reflect`). A role with no line keeps its default, and a role line of `inherit-parent` or `auto` runs that role on the parent chat model (omit `model`). Each code playbook passes only the model id from its line (`feature, refactoring`, `bug-fix`, `perf-issue`, or `hillclimb`), validated against `grok models`. The id is `grok-4.6` in `grok-4.6 xhigh`. The hardest changes read `hardest tasks`. Prose and judgment read `judgment and prose`. When `spawn_subagent` has no `model` argument, omit `model` for this session so every role inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. -**Match the model to the job.** Before each spawn, start from the role's line or default and pick the reasoning level the job needs. The `/setup-pstack` budget sets each role's starting level, xhigh by default. Raise it one step, up to max, for the hardest or subtlest work. Lower it one step for judgment and two or more steps for routine work. Verification is the exception and stays on the `swarm workers` model (`grok-4.6 xhigh` by default). Change only the effort and keep the model the role names, so a diverse panel stays diverse. If the model does not offer that effort, use its nearest offered effort. A role on `auto` or `inherit-parent` stays on the parent model. Set the effort only when `spawn_subagent` lists an effort argument. If the `spawn_subagent` tool rejects the adjusted slug, do not try another adjusted slug. Use the role's line or default and say so. If `spawn_subagent` rejects that too, continue with the calling skill's fallback. A slug counts as rejected once a `spawn_subagent` error says that the model is not valid, or leaves the slug out of its list of valid slugs. Never pass a rejected slug, and skip any fallback whose slug is already rejected. When the calling skill has no fallback, use the closest slug from `grok models`, or omit `model` so the subagent runs on the parent model when the list has none. When the calling skill's fallback runs out of slugs without saying what to do next, omit `model` too. Say so in both cases. +**Match the model to the job.** Before each spawn, start from the role's line or default and pick the reasoning level the job needs. The `/setup-pstack` budget sets each role's starting level, xhigh by default. Raise it one step, up to max, for the hardest or subtlest work. Lower it one step for judgment and two or more steps for routine work. Verification is the exception and stays on the `swarm workers` model id (`grok-4.6` at `xhigh` by default). Change only the effort and keep the model the role names, so a diverse panel stays diverse. If the model does not offer that effort, use its nearest offered effort. A role on `auto` or `inherit-parent` stays on the parent model. Set the effort only when `spawn_subagent` lists an effort argument. This CLI does not, so the child keeps the parent's effort. If the `spawn_subagent` tool rejects the adjusted slug, do not try another adjusted slug. Use the model id from the role's line or default and say so. If `spawn_subagent` rejects that too, continue with the calling skill's fallback. A slug counts as rejected once a `spawn_subagent` error says that the model is not valid, or leaves the slug out of its list of valid slugs. Never pass a rejected slug, and skip any fallback whose slug is already rejected. When the calling skill has no fallback, use the closest slug from `grok models`, or omit `model` so the subagent runs on the parent model when the list has none. When the calling skill's fallback runs out of slugs without saying what to do next, omit `model` too. Say so in both cases. You own every subagent's work. Review the diff and write your own summary, don't pass through what it said. A second opinion is the same prompt against a different model. Agreement is high-signal. diff --git a/skills/reflect/SKILL.md b/skills/reflect/SKILL.md index 7ae159e..40f91a5 100644 --- a/skills/reflect/SKILL.md +++ b/skills/reflect/SKILL.md @@ -25,7 +25,7 @@ Read the first lines of each candidate. Take the file whose opening user message One message, three `spawn_subagent` calls, omit `subagent_type`, the default tool set. Do not pass a readonly flag. Reviewers need MCP, which stays available through `search_tool` and `use_tool`. -Each reviewer and the synthesizer name a role line in the `pstack-models.md` rule and a default. Set `model` to that line's value, or to the default if the rule or the line is missing. Leave `model` unset when the value is `auto` or `inherit-parent`. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. +Each reviewer and the synthesizer name a role line in the `pstack-models.md` rule and a default. Set `model` to the model id in that line's value (`grok-4.6` from `grok-4.6 xhigh`), or to the default model id if the rule or the line is missing. Validate the id against `grok models`. Leave `model` unset when the value is `auto` or `inherit-parent`. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. | Lens | Role line | Default `model` | Prompt template | |---|---|---|---| diff --git a/skills/setup-pstack/SKILL.md b/skills/setup-pstack/SKILL.md index 1ff9e9f..5b0bf7f 100644 --- a/skills/setup-pstack/SKILL.md +++ b/skills/setup-pstack/SKILL.md @@ -13,7 +13,7 @@ Write `~/.grok/rules/pstack-models.md`. Every markdown file in `~/.grok/rules/` Run `grok models` and read the `spawn_subagent` schema. The slugs `grok models` lists are the only real slugs you may write. Read each model's effort levels from `grok models` / `/model`. If you cannot detect any slug, ask the user to paste the slugs they have access to. Never write a real slug you have not confirmed. The aliases `inherit-parent` and `auto` are always valid even though they are not detected slugs. Both mean omit `model`. -When `spawn_subagent` lists `model`, pass it. When it does not, this session inherits the parent. Expose the argument for the next session before you finish setup: +When `spawn_subagent` lists `model`, pass only the model id (`grok-4.6` from a `grok-4.6 xhigh` line), validated against `grok models`. When it does not, this session inherits the parent. Expose the argument for the next session before you finish setup: - Leave the user config alone when `GROK_SUBAGENT_MODEL_INHERITANCE` is set, when `GROK_CONFIG` or `GROK_CONFIG_PATH` sets `features.subagent_model_inheritance`, or when `$GROK_HOME/requirements.toml` (default `~/.grok/requirements.toml`), `/etc/grok/requirements.toml`, or an MDM profile pins that key. Name the pin in the confirmation. This session omits `model`. - Otherwise set `subagent_model_inheritance = false` under `[features]` in `$GROK_HOME/config.toml` (default `~/.grok/config.toml`). Preserve every other key and table. An explicit false overrides the remote default. The flag is read at session start, so this session still omits `model`. @@ -39,7 +39,7 @@ The default role-to-model mapping is the rule shape shown in step 5 below. If `~ ### 4. Validate -Every real slug written must be in the detected set. `inherit-parent` and `auto` always pass. If a chosen real slug is not available, stop and ask again. +Every real model id written must be in the detected set. The effort word after the id is not a slug. `inherit-parent` and `auto` always pass. If a chosen real model id is not available, stop and ask again. ### 5. Write the rule diff --git a/skills/swarm/SKILL.md b/skills/swarm/SKILL.md index e3b124b..f1209d3 100644 --- a/skills/swarm/SKILL.md +++ b/skills/swarm/SKILL.md @@ -19,7 +19,7 @@ Open a todolist with one entry per phase before launching anything. 1. State the done predicate and the artifact or report the swarm must return. 2. Choose the shape. Partition into slices, race N workers on identical briefs, or mix both. For a race or mixed shape, declare `first pass`, `rank all`, or `best-of` before spawning. -3. Pick the worker model from the `swarm workers` line in `~/.grok/rules/pstack-models.md`. If the rule or that line is missing, use `grok-4.6` at `xhigh`. For `auto` or `inherit-parent`, omit `model` so the workers run on the parent model. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. For a model race, name each arm's model up front. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. +3. Pick the worker model id from the `swarm workers` line in `~/.grok/rules/pstack-models.md` (`grok-4.6` from `grok-4.6 xhigh`), and validate it against `grok models`. If the rule or that line is missing, use `grok-4.6` at `xhigh`. For `auto` or `inherit-parent`, omit `model` so the workers run on the parent model. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. For a model race, name each arm's model up front. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. 4. Give each worker its own writable output when it writes. When workers verify or measure commits, each brief names the exact SHAs. A measurement brief also names the method (sample count, what one sample is, order). The worker records both in its result. ## Phase B: Fan out diff --git a/skills/why/SKILL.md b/skills/why/SKILL.md index 07688f9..3af91e3 100644 --- a/skills/why/SKILL.md +++ b/skills/why/SKILL.md @@ -8,7 +8,7 @@ disable-model-invocation: true Investigate the motivation and intent behind code. The companion `how` skill answers what the code does and how it works. `why` answers what forces led to its shape. -Each spawn below names a role line in the `pstack-models.md` rule and a default. Set `model` to that line's value, or to the default if the rule or the line is missing. Leave `model` unset when the value is `auto` or `inherit-parent`. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. +Each spawn below names a role line in the `pstack-models.md` rule and a default. Set `model` to the model id in that line's value (`grok-4.6` from `grok-4.6 xhigh`), or to the default model id if the rule or the line is missing. Validate the id against `grok models`. Leave `model` unset when the value is `auto` or `inherit-parent`. If `spawn_subagent` has no `model` argument, omit it for this session so the child inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. If it rejects a slug, use the closest slug from `grok models` and say so. Pick each subagent's reasoning effort for the job, per the Subagents section of the **poteto-mode** skill. Every spawn omits `subagent_type`. There is no readonly flag. Investigators must not write files. Say that in the prompt. MCP stays available through `search_tool` and `use_tool`, which the investigators and the synthesizer's citation spot-checks need. From f5a8dbf71b94c9f15e405272a02e05028de25e51 Mon Sep 17 00:00:00 2001 From: Cursor Agent Date: Mon, 9 Nov 2026 03:05:08 +0000 Subject: [PATCH 2/2] Keep the hardest-tasks clause on the playbook model line. Pass only the model id, and keep upstream's ", and the hardest changes read `hardest tasks`" wording. Co-authored-by: Kaitian Xie --- skills/poteto-mode/SKILL.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/skills/poteto-mode/SKILL.md b/skills/poteto-mode/SKILL.md index 63391a7..2721a76 100644 --- a/skills/poteto-mode/SKILL.md +++ b/skills/poteto-mode/SKILL.md @@ -93,7 +93,7 @@ Read the leaf skill in full for any principle you apply. Each entry names when i **Use `subagent_type: "pstack:poteto-agent"` for any subagent you spawn inside a playbook step** (code-writing delegates, ad-hoc helpers). The chat that receives `/poteto-mode` runs the playbook and is the only chat that may call `spawn_subagent`. A child cannot spawn. Never spawn a `poteto-agent` to run the playbook itself. Routed workflow skills (`how`, `why`, `interrogate`, `reflect`, `swarm`) omit `subagent_type`, or set the type they name. Respect what the skill prescribes, don't override to `poteto-agent`. -**Defaults for every `spawn_subagent` call.** `run_in_background: true` (a steer or resume of a running agent too, because a foreground call holds your turn while other reports wait). There is no readonly flag. When a child must not edit files, say so in the prompt. MCP stays available through `search_tool` and `use_tool`. Pass file pointers, not inlined context. Pass `model` only when the tool schema includes it. Otherwise omit `model` so the child inherits the parent. If it rejects a slug, use the closest slug from `grok models` and say so. Defaults are `grok-4.6` at `xhigh` for code and for prose and judgment. Code delegates tier by difficulty. The hardest changes (cross-cutting design, gnarly concurrency, subtle algorithms) go to your strongest judgment model (`grok-4.6` at `xhigh`), whether the task needs judgment on vague intent or is a precisely specified sequence of steps to execute to the letter. Trivial mechanical edits go to your fast code model. Per-role lines in `~/.grok/rules/pstack-models.md` (written by `/setup-pstack`) override these defaults and the model choices in the routed skills (`how`, `why`, `arena`, `swarm`, `architect`, `interrogate`, `reflect`). A role with no line keeps its default, and a role line of `inherit-parent` or `auto` runs that role on the parent chat model (omit `model`). Each code playbook passes only the model id from its line (`feature, refactoring`, `bug-fix`, `perf-issue`, or `hillclimb`), validated against `grok models`. The id is `grok-4.6` in `grok-4.6 xhigh`. The hardest changes read `hardest tasks`. Prose and judgment read `judgment and prose`. When `spawn_subagent` has no `model` argument, omit `model` for this session so every role inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. +**Defaults for every `spawn_subagent` call.** `run_in_background: true` (a steer or resume of a running agent too, because a foreground call holds your turn while other reports wait). There is no readonly flag. When a child must not edit files, say so in the prompt. MCP stays available through `search_tool` and `use_tool`. Pass file pointers, not inlined context. Pass `model` only when the tool schema includes it. Otherwise omit `model` so the child inherits the parent. If it rejects a slug, use the closest slug from `grok models` and say so. Defaults are `grok-4.6` at `xhigh` for code and for prose and judgment. Code delegates tier by difficulty. The hardest changes (cross-cutting design, gnarly concurrency, subtle algorithms) go to your strongest judgment model (`grok-4.6` at `xhigh`), whether the task needs judgment on vague intent or is a precisely specified sequence of steps to execute to the letter. Trivial mechanical edits go to your fast code model. Per-role lines in `~/.grok/rules/pstack-models.md` (written by `/setup-pstack`) override these defaults and the model choices in the routed skills (`how`, `why`, `arena`, `swarm`, `architect`, `interrogate`, `reflect`). A role with no line keeps its default, and a role line of `inherit-parent` or `auto` runs that role on the parent chat model (omit `model`). Each code playbook passes only the model id from its line (`feature, refactoring`, `bug-fix`, `perf-issue`, or `hillclimb`), validated against `grok models`, and the hardest changes read `hardest tasks`. Prose and judgment read `judgment and prose`. When `spawn_subagent` has no `model` argument, omit `model` for this session so every role inherits the parent, and apply step 1 of **setup-pstack** so the next session can pass `model`. **Match the model to the job.** Before each spawn, start from the role's line or default and pick the reasoning level the job needs. The `/setup-pstack` budget sets each role's starting level, xhigh by default. Raise it one step, up to max, for the hardest or subtlest work. Lower it one step for judgment and two or more steps for routine work. Verification is the exception and stays on the `swarm workers` model id (`grok-4.6` at `xhigh` by default). Change only the effort and keep the model the role names, so a diverse panel stays diverse. If the model does not offer that effort, use its nearest offered effort. A role on `auto` or `inherit-parent` stays on the parent model. Set the effort only when `spawn_subagent` lists an effort argument. This CLI does not, so the child keeps the parent's effort. If the `spawn_subagent` tool rejects the adjusted slug, do not try another adjusted slug. Use the model id from the role's line or default and say so. If `spawn_subagent` rejects that too, continue with the calling skill's fallback. A slug counts as rejected once a `spawn_subagent` error says that the model is not valid, or leaves the slug out of its list of valid slugs. Never pass a rejected slug, and skip any fallback whose slug is already rejected. When the calling skill has no fallback, use the closest slug from `grok models`, or omit `model` so the subagent runs on the parent model when the list has none. When the calling skill's fallback runs out of slugs without saying what to do next, omit `model` too. Say so in both cases.