feat: model overlays with explicit --model flag (no auto-detect)

Adds a per-model behavioral patch layer orthogonal to the host axis.
Different LLMs have different tendencies (GPT won't stop, Gemini
over-explains, o-series wants structured output). Overlays nudge each
model toward better defaults for gstack workflows.

Codex review caught three landmines the prior reviews missed:
1. Host != model — Claude Code can run any Claude model, Codex runs
   GPT/o-series, Cursor fronts multiple providers. Auto-detecting from
   host would lie. Dropped auto-detect. --model is explicit (default
   claude). Missing overlay file → empty string (graceful).
2. Import cycle — putting Model in resolvers/types.ts would cycle
   through hosts/index. Created neutral scripts/models.ts instead.
3. "Final say" is dangerous — overlay at the end of preamble could
   override STOP points, AskUserQuestion gates, /ship review gates.
   Placed overlay after spawned-session-check but before voice + tier
   sections. Wrapper heading adds explicit subordination language on
   every overlay: "subordinate to skill workflow, STOP points,
   AskUserQuestion gates, plan-mode safety, and /ship review gates."

Changes:
- scripts/models.ts: new neutral module. ALL_MODEL_NAMES, Model type,
  resolveModel() for family heuristics (gpt-5.4-mini → gpt-5.4, o3 →
  o-series, claude-opus-4-7 → claude), validateModel() helper.
- scripts/resolvers/types.ts: import Model, add ctx.model field.
- scripts/resolvers/model-overlay.ts: new resolver. Reads
  model-overlays/{model}.md. Supports {{INHERIT:base}} directive at
  top of file for concat (gpt-5.4 inherits gpt). Cycle guard.
- scripts/resolvers/index.ts: register MODEL_OVERLAY resolver.
- scripts/resolvers/preamble.ts: wire generateModelOverlay into
  composition before voice. Print MODEL_OVERLAY: {model} in preamble
  bash so users can see which overlay is active. Filter empty sections.
- scripts/gen-skill-docs.ts: parse --model CLI flag. Default claude.
  Unknown model → throw with list of valid options.
- model-overlays/{claude,gpt,gpt-5.4,gemini,o-series}.md: behavioral
  patches per model family. gpt-5.4.md uses {{INHERIT:gpt}} to extend
  gpt.md without duplication.
- test/gen-skill-docs.test.ts: fix qa-only guardrail regex scope.
  Was matching Edit/Glob/Grep anywhere after `allowed-tools:` in the
  whole file. Now scoped to frontmatter only. Body prose (Claude
  overlay references Edit as a tool) correctly no longer breaks it.

Verification:
- bun run gen:skill-docs --host all --dry-run → all fresh
- bun run gen:skill-docs --model gpt-5.4 → concat works, gpt.md +
  gpt-5.4.md content appears in order
- bun run gen:skill-docs --model unknown → errors with valid list
- All generated skills contain MODEL_OVERLAY: claude in preamble
- Golden ship fixtures regenerated

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

SKILL.md

@@ -88,6 +88,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -255,6 +256,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 **Tone:** direct, concrete, sharp, never corporate, never academic. Sound like a builder, not a consultant. Name the file, the function, the command. No filler, no throat-clearing.

autoplan/SKILL.md

@@ -98,6 +98,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -265,6 +266,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

benchmark/SKILL.md

@@ -91,6 +91,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -258,6 +259,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 **Tone:** direct, concrete, sharp, never corporate, never academic. Sound like a builder, not a consultant. Name the file, the function, the command. No filler, no throat-clearing.

browse/SKILL.md

@@ -90,6 +90,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -257,6 +258,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 **Tone:** direct, concrete, sharp, never corporate, never academic. Sound like a builder, not a consultant. Name the file, the function, the command. No filler, no throat-clearing.

canary/SKILL.md

@@ -90,6 +90,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -257,6 +258,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

checkpoint/SKILL.md

@@ -93,6 +93,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -260,6 +261,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

codex/SKILL.md

@@ -92,6 +92,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -259,6 +260,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

cso/SKILL.md

@@ -95,6 +95,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -262,6 +263,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

design-consultation/SKILL.md

@@ -95,6 +95,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -262,6 +263,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

design-html/SKILL.md

@@ -97,6 +97,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -264,6 +265,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

design-review/SKILL.md

@@ -95,6 +95,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -262,6 +263,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

design-shotgun/SKILL.md

@@ -92,6 +92,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -259,6 +260,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

devex-review/SKILL.md

@@ -95,6 +95,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -262,6 +263,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

document-release/SKILL.md

@@ -92,6 +92,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -259,6 +260,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

health/SKILL.md

@@ -92,6 +92,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -259,6 +260,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

investigate/SKILL.md

@@ -107,6 +107,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -274,6 +275,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

land-and-deploy/SKILL.md

@@ -89,6 +89,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -256,6 +257,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

learn/SKILL.md

@@ -92,6 +92,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -259,6 +260,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

model-overlays/claude.md

@@ -0,0 +1,10 @@
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.

model-overlays/gemini.md

@@ -0,0 +1,10 @@
+**Conciseness constraint.** Keep non-code text output short. Aim for under 3 lines
+for routine responses unless the user explicitly asks for detail. Code blocks and
+command output do not count toward the limit.
+
+**Bias toward action.** Run commands and show results rather than explaining what
+commands you would run. The user sees the command and the output — they don't need
+narration.
+
+**Structured output when useful.** Tables, bullet points, and code blocks beat prose
+for lists of things. Prose is for explaining; structure is for presenting.

model-overlays/gpt-5.4.md

@@ -0,0 +1,15 @@
+{{INHERIT:gpt}}
+
+**Anti-verbosity protocol (additional).** Your default output mode is too verbose for
+tools that value terse output. Constrain:
+
+- Status updates: one line, not a paragraph.
+- Code explanations: only when the user asked for one, or when the code is genuinely
+  surprising.
+- Do not narrate what you are about to do. Just do it.
+- Do not repeat the user's request back to them.
+- When showing code changes, show the changed lines with minimal surrounding context.
+- Markdown headings are not decoration. Use them only when structural.
+
+**Cap answers at the shortest form that contains the answer.** If the answer is a
+one-line command, reply with a one-line command.

model-overlays/gpt.md

@@ -0,0 +1,14 @@
+**Completion bias.** Do not end your turn with a partial solution when the full
+solution is reachable. If you encounter an error, debug it. If a test fails, fix it.
+If something is ambiguous, make your best judgment and proceed — don't stop and ask
+unless you're genuinely blocked.
+
+**Prefer doing over listing.** When you'd be tempted to write "you could also try X,
+Y, or Z," try the best option yourself. Pick, execute, report results.
+
+**No preamble.** Skip "Great question!", "Let me help with that", and restating the
+user's request. Start with the work.
+
+**Reminder: subordination applies.** When a skill workflow says STOP, stop. When the
+skill asks via AskUserQuestion, that is the wait-for-user gate, not an ambiguity.
+Completion bias does not override safety gates.

model-overlays/o-series.md

@@ -0,0 +1,11 @@
+**Reasoning model behavior.** You have strong internal reasoning. Use it, but do not
+expose chain-of-thought in outputs unless the user asks to see your reasoning.
+Surface the conclusion plus evidence, not the reasoning chain.
+
+**Structured outputs preferred.** Tables or bullet points over prose paragraphs
+when presenting analysis. Prose is for explanation and context; structure is for
+findings, options, and comparisons.
+
+**Completion bias (subordinate to safety gates).** Do not stop with partial
+solutions when the full solution is reachable. But skill workflow STOP points,
+AskUserQuestion gates, and /ship review gates always win over completion bias.

office-hours/SKILL.md

@@ -99,6 +99,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -266,6 +267,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

open-gstack-browser/SKILL.md

@@ -89,6 +89,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -256,6 +257,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

pair-agent/SKILL.md

@@ -90,6 +90,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -257,6 +258,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

plan-ceo-review/SKILL.md

@@ -95,6 +95,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -262,6 +263,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

plan-design-review/SKILL.md

@@ -93,6 +93,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -260,6 +261,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

plan-devex-review/SKILL.md

@@ -97,6 +97,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -264,6 +265,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

plan-eng-review/SKILL.md

@@ -95,6 +95,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -262,6 +263,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

qa-only/SKILL.md

@@ -91,6 +91,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -258,6 +259,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

qa/SKILL.md

@@ -97,6 +97,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -264,6 +265,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

retro/SKILL.md

@@ -90,6 +90,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -257,6 +258,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

review/SKILL.md

@@ -93,6 +93,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -260,6 +261,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

scripts/gen-skill-docs.ts

@@ -43,6 +43,21 @@ const HOST_ARG_VAL: HostArg = (() => {
 // For single-host mode, HOST is the host. For --host all, it's set per iteration below.
 let HOST: Host = HOST_ARG_VAL === 'all' ? 'claude' : HOST_ARG_VAL;
 
+// ─── Model Overlay Selection ────────────────────────────────
+// --model is explicit. We do NOT auto-detect from host (host ≠ model).
+// Default is 'claude'. Missing overlay file → empty string (graceful).
+import { ALL_MODEL_NAMES, resolveModel, type Model } from './models';
+const MODEL_ARG = process.argv.find(a => a.startsWith('--model'));
+const MODEL_ARG_VAL: Model = (() => {
+  if (!MODEL_ARG) return 'claude';
+  const val = MODEL_ARG.includes('=') ? MODEL_ARG.split('=')[1] : process.argv[process.argv.indexOf(MODEL_ARG) + 1];
+  const resolved = resolveModel(val);
+  if (!resolved) {
+    throw new Error(`Unknown model: ${val}. Use ${ALL_MODEL_NAMES.join(', ')}, or a family variant (e.g., claude-opus-4-7, gpt-5.4-mini, o3).`);
+  }
+  return resolved;
+})();
+
 // HostPaths, HOST_PATHS, and TemplateContext imported from ./resolvers/types (line 7-8)
 // Design constants (AI_SLOP_BLACKLIST, OPENAI_HARD_REJECTIONS, OPENAI_LITMUS_CHECKS)
 // live in ./resolvers/constants and are consumed by resolvers directly.
@@ -398,7 +413,7 @@ function processTemplate(tmplPath: string, host: Host = 'claude'): { outputPath:
   const tierMatch = tmplContent.match(/^preamble-tier:\s*(\d+)$/m);
   const preambleTier = tierMatch ? parseInt(tierMatch[1], 10) : undefined;
 
-  const ctx: TemplateContext = { skillName, tmplPath, benefitsFrom, host, paths: HOST_PATHS[host], preambleTier };
+  const ctx: TemplateContext = { skillName, tmplPath, benefitsFrom, host, paths: HOST_PATHS[host], preambleTier, model: MODEL_ARG_VAL };
 
   // Replace placeholders (supports parameterized: {{NAME:arg1:arg2}})
   // Config-driven: suppressedResolvers return empty string for this host

scripts/models.ts

@@ -0,0 +1,68 @@
+/**
+ * Model taxonomy — neutral module with no imports from hosts/ or resolvers/.
+ *
+ * Model families supported by model overlays in model-overlays/{family}.md.
+ * Host configs can reference these as `defaultModel` strings (validated at
+ * generation time), but the model axis is independent of the host axis.
+ *
+ * IMPORTANT: host ≠ model. Claude Code can run any Claude model (Opus, Sonnet,
+ * Haiku, future). Codex CLI runs GPT/o-series models. Cursor and OpenCode can
+ * front multiple providers. We do NOT auto-detect the model from the host —
+ * users pass --model explicitly. Default is 'claude'.
+ */
+
+export const ALL_MODEL_NAMES = [
+  'claude',
+  'gpt',
+  'gpt-5.4',
+  'gemini',
+  'o-series',
+] as const;
+
+export type Model = (typeof ALL_MODEL_NAMES)[number];
+
+/**
+ * Resolve a model argument from CLI input to a known Model family.
+ *
+ * Precedence rules:
+ * 1. Exact match against ALL_MODEL_NAMES → return as-is.
+ * 2. Family heuristics for common variants:
+ *    - `gpt-5.4-mini`, `gpt-5.4-turbo`, `gpt-5.4-*` → `gpt-5.4`
+ *    - `gpt-*` (anything else GPT) → `gpt`
+ *    - `o3`, `o4`, `o4-mini`, `o1`, `o1-mini`, `o1-pro` → `o-series`
+ *    - `claude-*` (sonnet, opus, haiku, any version) → `claude`
+ *    - `gemini-*` (2.5-pro, flash, etc.) → `gemini`
+ * 3. Unknown input → returns null (caller decides: error, or fall back).
+ *
+ * The resolver file in model-overlays/{model}.md applies further fallback
+ * (e.g., missing gpt-5.4.md falls back to gpt.md). This function only
+ * normalizes CLI input to a family name.
+ */
+export function resolveModel(input: string): Model | null {
+  const s = input.trim();
+  if (!s) return null;
+
+  // Exact match first
+  if ((ALL_MODEL_NAMES as readonly string[]).includes(s)) {
+    return s as Model;
+  }
+
+  // Family heuristics
+  if (/^gpt-5\.4(-|$)/.test(s)) return 'gpt-5.4';
+  if (/^gpt(-|$)/.test(s)) return 'gpt';
+  if (/^o[0-9]+(-|$)/.test(s)) return 'o-series';
+  if (/^claude(-|$)/.test(s)) return 'claude';
+  if (/^gemini(-|$)/.test(s)) return 'gemini';
+
+  return null;
+}
+
+/**
+ * Validate a string against ALL_MODEL_NAMES. Used by host-config validators
+ * when a HostConfig declares `defaultModel`. Returns an error message or null
+ * if valid.
+ */
+export function validateModel(input: string): string | null {
+  if ((ALL_MODEL_NAMES as readonly string[]).includes(input)) return null;
+  return `'${input}' is not a known model. Use ${ALL_MODEL_NAMES.join(', ')}.`;
+}

scripts/resolvers/index.ts

@@ -18,6 +18,7 @@ import { generateConfidenceCalibration } from './confidence';
 import { generateInvokeSkill } from './composition';
 import { generateReviewArmy } from './review-army';
 import { generateDxFramework } from './dx';
+import { generateModelOverlay } from './model-overlay';
 
 export const RESOLVERS: Record<string, ResolverFn> = {
   SLUG_EVAL: generateSlugEval,
@@ -62,4 +63,5 @@ export const RESOLVERS: Record<string, ResolverFn> = {
   REVIEW_ARMY: generateReviewArmy,
   CROSS_REVIEW_DEDUP: generateCrossReviewDedup,
   DX_FRAMEWORK: generateDxFramework,
+  MODEL_OVERLAY: generateModelOverlay,
 };

scripts/resolvers/model-overlay.ts

@@ -0,0 +1,60 @@
+/**
+ * Model overlay resolver — reads model-overlays/{model}.md and returns it
+ * wrapped in a subordinate behavioral-patch section.
+ *
+ * Precedence:
+ *   1. Exact match: ctx.model === 'gpt-5.4' → reads model-overlays/gpt-5.4.md
+ *   2. INHERIT directive: if the file's first non-whitespace line is
+ *      `{{INHERIT:claude}}`, the resolver reads model-overlays/claude.md first
+ *      and concatenates it ahead of the rest of this file's content.
+ *      This lets `gpt-5.4.md` build on top of `gpt.md` without duplication.
+ *   3. Missing file: returns empty string (graceful degradation, no error).
+ *   4. No ctx.model set: returns empty string.
+ *
+ * The returned block is subordinate to skill workflow, safety gates, and
+ * AskUserQuestion instructions. The subordination language is part of the
+ * wrapper heading so it appears with every overlay regardless of file content.
+ */
+
+import * as fs from 'fs';
+import * as path from 'path';
+import type { TemplateContext } from './types';
+
+const OVERLAY_DIR = path.resolve(import.meta.dir, '../../model-overlays');
+
+const INHERIT_RE = /^\s*\{\{INHERIT:([a-z0-9-]+(?:\.[0-9]+)*)\}\}\s*\n/;
+
+function readOverlay(model: string, seen: Set<string> = new Set()): string {
+  if (seen.has(model)) return ''; // cycle guard
+  seen.add(model);
+
+  const filePath = path.join(OVERLAY_DIR, `${model}.md`);
+  if (!fs.existsSync(filePath)) return '';
+
+  const raw = fs.readFileSync(filePath, 'utf-8');
+  const match = raw.match(INHERIT_RE);
+  if (!match) return raw.trim();
+
+  const baseModel = match[1];
+  const base = readOverlay(baseModel, seen);
+  const rest = raw.replace(INHERIT_RE, '').trim();
+
+  if (!base) return rest;
+  return `${base}\n\n${rest}`;
+}
+
+export function generateModelOverlay(ctx: TemplateContext): string {
+  if (!ctx.model) return '';
+
+  const content = readOverlay(ctx.model);
+  if (!content) return '';
+
+  return `## Model-Specific Behavioral Patch (${ctx.model})
+
+The following nudges are tuned for the ${ctx.model} model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+${content}`;
+}

scripts/resolvers/preamble.ts

@@ -1,5 +1,6 @@
 import type { TemplateContext } from './types';
 import { getHostConfig } from '../../hosts/index';
+import { generateModelOverlay } from './model-overlay';
 
 /**
  * Preamble architecture — why every skill needs this
@@ -97,6 +98,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: ${ctx.model ?? 'none'}"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 \`\`\``;
@@ -747,10 +749,11 @@ export function generatePreamble(ctx: TemplateContext): string {
     generateRoutingInjection(ctx),
     generateVendoringDeprecation(ctx),
     generateSpawnedSessionCheck(),
+    generateModelOverlay(ctx),
     generateVoiceDirective(tier),
     ...(tier >= 2 ? [generateContextRecovery(ctx), generateAskUserFormat(ctx), generateCompletenessSection(), generateContextHealth()] : []),
     ...(tier >= 3 ? [generateRepoModeSection(), generateSearchBeforeBuildingSection(ctx)] : []),
     generateCompletionStatus(ctx),
   ];
-  return sections.join('\n\n');
+  return sections.filter(s => s && s.trim().length > 0).join('\n\n');
 }

scripts/resolvers/types.ts

@@ -47,6 +47,9 @@ function buildHostPaths(): Record<string, HostPaths> {
 
 export const HOST_PATHS: Record<string, HostPaths> = buildHostPaths();
 
+import type { Model } from '../models';
+export type { Model } from '../models';
+
 export interface TemplateContext {
   skillName: string;
   tmplPath: string;
@@ -54,6 +57,7 @@ export interface TemplateContext {
   host: Host;
   paths: HostPaths;
   preambleTier?: number;  // 1-4, controls which preamble sections are included
+  model?: Model;  // model family for behavioral overlay. Omitted/undefined → no overlay.
 }
 
 /** Resolver function signature. args is populated for parameterized placeholders like {{INVOKE_SKILL:name}}. */

setup-browser-cookies/SKILL.md

@@ -87,6 +87,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -254,6 +255,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 **Tone:** direct, concrete, sharp, never corporate, never academic. Sound like a builder, not a consultant. Name the file, the function, the command. No filler, no throat-clearing.

setup-deploy/SKILL.md

@@ -93,6 +93,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -260,6 +261,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

ship/SKILL.md

@@ -94,6 +94,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -261,6 +262,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

test/fixtures/golden/claude-ship-SKILL.md

@@ -94,6 +94,7 @@ if [ -d ".claude/skills/gstack" ] && [ ! -L ".claude/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -261,6 +262,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

test/fixtures/golden/codex-ship-SKILL.md

@@ -88,6 +88,7 @@ if [ -d ".agents/skills/gstack" ] && [ ! -L ".agents/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -255,6 +256,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

test/fixtures/golden/factory-ship-SKILL.md

@@ -90,6 +90,7 @@ if [ -d ".factory/skills/gstack" ] && [ ! -L ".factory/skills/gstack" ]; then
   fi
 fi
 echo "VENDORED_GSTACK: $_VENDORED"
+echo "MODEL_OVERLAY: claude"
 # Detect spawned session (OpenClaw or other orchestrator)
 [ -n "$OPENCLAW_SESSION" ] && echo "SPAWNED_SESSION: true" || true
 ```
@@ -257,6 +258,24 @@ AI orchestrator (e.g., OpenClaw). In spawned sessions:
 - Focus on completing the task and reporting results via prose output.
 - End with a completion report: what shipped, decisions made, anything uncertain.
 
+## Model-Specific Behavioral Patch (claude)
+
+The following nudges are tuned for the claude model family. They are
+**subordinate** to skill workflow, STOP points, AskUserQuestion gates, plan-mode
+safety, and /ship review gates. If a nudge below conflicts with skill instructions,
+the skill wins. Treat these as preferences, not rules.
+
+**Todo-list discipline.** When working through a multi-step plan, mark each task
+complete individually as you finish it. Do not batch-complete at the end. If a task
+turns out to be unnecessary, mark it skipped with a one-line reason.
+
+**Think before heavy actions.** For complex operations (refactors, migrations,
+non-trivial new features), briefly state your approach before executing. This lets
+the user course-correct cheaply instead of mid-flight.
+
+**Dedicated tools over Bash.** Prefer Read, Edit, Write, Glob, Grep over shell
+equivalents (cat, sed, find, grep). The dedicated tools are cheaper and clearer.
+
 ## Voice
 
 You are GStack, an open source AI builder framework shaped by Garry Tan's product, startup, and engineering judgment. Encode how he thinks, not his biography.

test/gen-skill-docs.test.ts

@@ -358,10 +358,17 @@ describe('gen-skill-docs', () => {
     const qaOnlyContent = fs.readFileSync(path.join(ROOT, 'qa-only', 'SKILL.md'), 'utf-8');
     expect(qaOnlyContent).toContain('Never fix bugs');
     expect(qaOnlyContent).toContain('NEVER fix anything');
-    // Should not have Edit, Glob, or Grep in allowed-tools
+    // Should not have Edit, Glob, or Grep in allowed-tools.
-    expect(qaOnlyContent).not.toMatch(/allowed-tools:[\s\S]*?Edit/);
+    // Scope to frontmatter (between the first two --- lines) — the body can
-    expect(qaOnlyContent).not.toMatch(/allowed-tools:[\s\S]*?Glob/);
+    // legitimately mention these tool names in prose (e.g., Claude model
-    expect(qaOnlyContent).not.toMatch(/allowed-tools:[\s\S]*?Grep/);
+    // overlay says "prefer Read, Edit, Write, Glob, Grep over Bash").
+    const fmMatch = qaOnlyContent.match(/^---\n([\s\S]*?)\n---/);
+    expect(fmMatch).not.toBeNull();
+    const frontmatter = fmMatch![1];
+    expect(frontmatter).toMatch(/allowed-tools:/);
+    expect(frontmatter).not.toMatch(/allowed-tools:[\s\S]*?- Edit/);
+    expect(frontmatter).not.toMatch(/allowed-tools:[\s\S]*?- Glob/);
+    expect(frontmatter).not.toMatch(/allowed-tools:[\s\S]*?- Grep/);
   });
 
   test('qa has fix-loop tools and phases', () => {