feat(gstack2): infer execution profiles

2026-09-21 12:20:48 +02:00 · 2026-07-20 16:22:43 -07:00
parent 9b5ae4071d
commit a1b2b05a18
20 changed files with 329 additions and 30 deletions
@@ -15,7 +15,7 @@ Before any substantive output, print these exact labels in this exact order. Res
 ```text
 Target: <concrete repository, product, URL, device, PR, or artifact>
 Mode: <selected top-level mode>
-Depth: <quick, standard, or deep>
+Depth: <readiness, standard, or deep>
 Mutation: <report-only or exact authorized mutation boundary>
 Active modules: <comma-separated internal specialist modules>
 Skipped modules: <comma-separated non-active mandatory modules with compact reasons>
@@ -27,7 +27,7 @@ Web context: <none, optional, local-browser, or production>
 1. Infer the mode from product stage, surface, requested artifact, mutation authorization, evidence needs, and deployment state. Do not route by keyword alone.
 2. Refine the public mode to the smallest applicable internal specialist set, then print the required execution header before any substantive output.
 3. Read each active module in full from the path shown in the mode/alias tables. Its specialist body, behavioral contract, STOP gates, and appended upstream judgment ports are binding. Read a lazy specialist phase in full only when the workflow reaches its package-local reference.
-4. Read `references/SHARED-JUDGMENT.md` and `references/AUTHORITY-POLICY.md` for every invocation. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
+4. Read `references/EXECUTION-PROFILES.md`, `references/SHARED-JUDGMENT.md`, and `references/AUTHORITY-POLICY.md` for every invocation. Infer Depth from structured operating conditions, then obey its mandatory modules, legal skips, artifacts, and claim limits. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
 5. If an old asset path is unavailable, use `references/ASSETS.md`. If legacy prose invokes another retired skill, resolve it through `references/COMPATIBILITY.md` and stay inside these six dispatchers.
 6. Preserve report-only versus mutation boundaries. Missing mutation authorization fails closed: do not edit merely because a specialist can fix. Commits, pushes, PRs, merges, deploys, messages, and other external mutations still require affirmative authority from the user.
 7. Match the user's language. Keep code identifiers, commands, and source quotations original when translation would reduce accuracy.
@@ -0,0 +1,28 @@
+<!-- GENERATED by scripts/gstack2/generate-skill-tree.ts; do not edit. -->
+# Inferred execution profiles
+
+Choose a profile from product stage, mutation authority, risk/evidence needs, and deployment state. Prompt keywords and a request to “be quick” are not routing evidence. A profile narrows or strengthens evidence; it never overrides a specialist’s binding question order, pressure, gates, mutation boundary, or exit behavior.
+
+## Smoke/readiness
+
+- Infer when: A narrow, reversible, pre-deployment or operational readiness decision needs bounded evidence; the requested claim is readiness, not completeness.
+- Mandatory modules: Every selected specialist module remains mandatory. Run its entry gates and the smallest source-authorized evidence path that can answer the readiness question.
+- Legal skips: Only source-declared conditional work whose condition is demonstrably false, or evidence unavailable after a named attempt. Never skip a STOP gate, approval boundary, reproduction/root-cause gate, or required physical-device/production evidence.
+- Artifacts: A readiness record naming the exact scope, probes run, evidence and freshness, failures, skipped work with reasons, and the next standard/deep step.
+- Claims: Only ready/not-ready for the named bounded decision. Must say “Readiness profile — not a complete review.” Never claim comprehensive, fully verified, production-safe, or no issues found outside the inspected scope.
+
+## Standard
+
+- Infer when: Normal feature or change work has bounded scope and risk and needs the selected specialist’s complete default workflow.
+- Mandatory modules: Every selected specialist module and all of its mandatory phases, gates, artifacts, and exit checks.
+- Legal skips: Only smart skips explicitly authorized by the specialist and supported by inspected evidence; list each skipped primary module and each skipped conditional phase.
+- Artifacts: Every artifact required by the selected specialist, plus evidence provenance, unresolved decisions, and a skip ledger.
+- Claims: Complete only for the named specialist scope and evidence layer. Broader product, security, production, or device claims require those modules and evidence.
+
+## Deep
+
+- Infer when: Risk, ambiguity, blast radius, cross-system effects, irreversible mutation, security/reliability needs, or production deployment demands stronger evidence.
+- Mandatory modules: Every selected specialist module, all mandatory phases and outside-voice/cross-consumer modules selected by the dispatcher, with unchanged STOP and approval gates.
+- Legal skips: Only specialist-authorized smart skips proven irrelevant. Missing, stale, malformed, or contradictory evidence is a gap or blocker, never a successful skip.
+- Artifacts: All specialist artifacts plus changed-input/unchanged-consumer trace, negative and failure-path evidence, provenance/freshness, rollback or reversibility evidence where applicable, and unresolved-risk ledger.
+- Claims: Complete only across the explicitly listed modules and evidence layers. “Confirmed” requires independent supporting evidence; production/device/security claims require matching production/device/security evidence.
@@ -15,7 +15,7 @@ Before any substantive output, print these exact labels in this exact order. Res
 ```text
 Target: <concrete repository, product, URL, device, PR, or artifact>
 Mode: <selected top-level mode>
-Depth: <quick, standard, or deep>
+Depth: <readiness, standard, or deep>
 Mutation: <report-only or exact authorized mutation boundary>
 Active modules: <comma-separated internal specialist modules>
 Skipped modules: <comma-separated non-active mandatory modules with compact reasons>
@@ -27,7 +27,7 @@ Web context: <none, optional, local-browser, or production>
 1. Infer the mode from product stage, surface, requested artifact, mutation authorization, evidence needs, and deployment state. Do not route by keyword alone.
 2. Refine the public mode to the smallest applicable internal specialist set, then print the required execution header before any substantive output.
 3. Read each active module in full from the path shown in the mode/alias tables. Its specialist body, behavioral contract, STOP gates, and appended upstream judgment ports are binding. Read a lazy specialist phase in full only when the workflow reaches its package-local reference.
-4. Read `references/SHARED-JUDGMENT.md` and `references/AUTHORITY-POLICY.md` for every invocation. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
+4. Read `references/EXECUTION-PROFILES.md`, `references/SHARED-JUDGMENT.md`, and `references/AUTHORITY-POLICY.md` for every invocation. Infer Depth from structured operating conditions, then obey its mandatory modules, legal skips, artifacts, and claim limits. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
 5. If an old asset path is unavailable, use `references/ASSETS.md`. If legacy prose invokes another retired skill, resolve it through `references/COMPATIBILITY.md` and stay inside these six dispatchers.
 6. Preserve report-only versus mutation boundaries. Missing mutation authorization fails closed: do not edit merely because a specialist can fix. Commits, pushes, PRs, merges, deploys, messages, and other external mutations still require affirmative authority from the user.
 7. Match the user's language. Keep code identifiers, commands, and source quotations original when translation would reduce accuracy.
@@ -0,0 +1,28 @@
+<!-- GENERATED by scripts/gstack2/generate-skill-tree.ts; do not edit. -->
+# Inferred execution profiles
+
+Choose a profile from product stage, mutation authority, risk/evidence needs, and deployment state. Prompt keywords and a request to “be quick” are not routing evidence. A profile narrows or strengthens evidence; it never overrides a specialist’s binding question order, pressure, gates, mutation boundary, or exit behavior.
+
+## Smoke/readiness
+
+- Infer when: A narrow, reversible, pre-deployment or operational readiness decision needs bounded evidence; the requested claim is readiness, not completeness.
+- Mandatory modules: Every selected specialist module remains mandatory. Run its entry gates and the smallest source-authorized evidence path that can answer the readiness question.
+- Legal skips: Only source-declared conditional work whose condition is demonstrably false, or evidence unavailable after a named attempt. Never skip a STOP gate, approval boundary, reproduction/root-cause gate, or required physical-device/production evidence.
+- Artifacts: A readiness record naming the exact scope, probes run, evidence and freshness, failures, skipped work with reasons, and the next standard/deep step.
+- Claims: Only ready/not-ready for the named bounded decision. Must say “Readiness profile — not a complete review.” Never claim comprehensive, fully verified, production-safe, or no issues found outside the inspected scope.
+
+## Standard
+
+- Infer when: Normal feature or change work has bounded scope and risk and needs the selected specialist’s complete default workflow.
+- Mandatory modules: Every selected specialist module and all of its mandatory phases, gates, artifacts, and exit checks.
+- Legal skips: Only smart skips explicitly authorized by the specialist and supported by inspected evidence; list each skipped primary module and each skipped conditional phase.
+- Artifacts: Every artifact required by the selected specialist, plus evidence provenance, unresolved decisions, and a skip ledger.
+- Claims: Complete only for the named specialist scope and evidence layer. Broader product, security, production, or device claims require those modules and evidence.
+
+## Deep
+
+- Infer when: Risk, ambiguity, blast radius, cross-system effects, irreversible mutation, security/reliability needs, or production deployment demands stronger evidence.
+- Mandatory modules: Every selected specialist module, all mandatory phases and outside-voice/cross-consumer modules selected by the dispatcher, with unchanged STOP and approval gates.
+- Legal skips: Only specialist-authorized smart skips proven irrelevant. Missing, stale, malformed, or contradictory evidence is a gap or blocker, never a successful skip.
+- Artifacts: All specialist artifacts plus changed-input/unchanged-consumer trace, negative and failure-path evidence, provenance/freshness, rollback or reversibility evidence where applicable, and unresolved-risk ledger.
+- Claims: Complete only across the explicitly listed modules and evidence layers. “Confirmed” requires independent supporting evidence; production/device/security claims require matching production/device/security evidence.
@@ -15,7 +15,7 @@ Before any substantive output, print these exact labels in this exact order. Res
 ```text
 Target: <concrete repository, product, URL, device, PR, or artifact>
 Mode: <selected top-level mode>
-Depth: <quick, standard, or deep>
+Depth: <readiness, standard, or deep>
 Mutation: <report-only or exact authorized mutation boundary>
 Active modules: <comma-separated internal specialist modules>
 Skipped modules: <comma-separated non-active mandatory modules with compact reasons>
@@ -27,7 +27,7 @@ Web context: <none, optional, local-browser, or production>
 1. Infer the mode from product stage, surface, requested artifact, mutation authorization, evidence needs, and deployment state. Do not route by keyword alone.
 2. Refine the public mode to the smallest applicable internal specialist set, then print the required execution header before any substantive output.
 3. Read each active module in full from the path shown in the mode/alias tables. Its specialist body, behavioral contract, STOP gates, and appended upstream judgment ports are binding. Read a lazy specialist phase in full only when the workflow reaches its package-local reference.
-4. Read `references/SHARED-JUDGMENT.md` and `references/AUTHORITY-POLICY.md` for every invocation. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
+4. Read `references/EXECUTION-PROFILES.md`, `references/SHARED-JUDGMENT.md`, and `references/AUTHORITY-POLICY.md` for every invocation. Infer Depth from structured operating conditions, then obey its mandatory modules, legal skips, artifacts, and claim limits. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
 5. If an old asset path is unavailable, use `references/ASSETS.md`. If legacy prose invokes another retired skill, resolve it through `references/COMPATIBILITY.md` and stay inside these six dispatchers.
 6. Preserve report-only versus mutation boundaries. Missing mutation authorization fails closed: do not edit merely because a specialist can fix. Commits, pushes, PRs, merges, deploys, messages, and other external mutations still require affirmative authority from the user.
 7. Match the user's language. Keep code identifiers, commands, and source quotations original when translation would reduce accuracy.
@@ -0,0 +1,28 @@
+<!-- GENERATED by scripts/gstack2/generate-skill-tree.ts; do not edit. -->
+# Inferred execution profiles
+
+Choose a profile from product stage, mutation authority, risk/evidence needs, and deployment state. Prompt keywords and a request to “be quick” are not routing evidence. A profile narrows or strengthens evidence; it never overrides a specialist’s binding question order, pressure, gates, mutation boundary, or exit behavior.
+
+## Smoke/readiness
+
+- Infer when: A narrow, reversible, pre-deployment or operational readiness decision needs bounded evidence; the requested claim is readiness, not completeness.
+- Mandatory modules: Every selected specialist module remains mandatory. Run its entry gates and the smallest source-authorized evidence path that can answer the readiness question.
+- Legal skips: Only source-declared conditional work whose condition is demonstrably false, or evidence unavailable after a named attempt. Never skip a STOP gate, approval boundary, reproduction/root-cause gate, or required physical-device/production evidence.
+- Artifacts: A readiness record naming the exact scope, probes run, evidence and freshness, failures, skipped work with reasons, and the next standard/deep step.
+- Claims: Only ready/not-ready for the named bounded decision. Must say “Readiness profile — not a complete review.” Never claim comprehensive, fully verified, production-safe, or no issues found outside the inspected scope.
+
+## Standard
+
+- Infer when: Normal feature or change work has bounded scope and risk and needs the selected specialist’s complete default workflow.
+- Mandatory modules: Every selected specialist module and all of its mandatory phases, gates, artifacts, and exit checks.
+- Legal skips: Only smart skips explicitly authorized by the specialist and supported by inspected evidence; list each skipped primary module and each skipped conditional phase.
+- Artifacts: Every artifact required by the selected specialist, plus evidence provenance, unresolved decisions, and a skip ledger.
+- Claims: Complete only for the named specialist scope and evidence layer. Broader product, security, production, or device claims require those modules and evidence.
+
+## Deep
+
+- Infer when: Risk, ambiguity, blast radius, cross-system effects, irreversible mutation, security/reliability needs, or production deployment demands stronger evidence.
+- Mandatory modules: Every selected specialist module, all mandatory phases and outside-voice/cross-consumer modules selected by the dispatcher, with unchanged STOP and approval gates.
+- Legal skips: Only specialist-authorized smart skips proven irrelevant. Missing, stale, malformed, or contradictory evidence is a gap or blocker, never a successful skip.
+- Artifacts: All specialist artifacts plus changed-input/unchanged-consumer trace, negative and failure-path evidence, provenance/freshness, rollback or reversibility evidence where applicable, and unresolved-risk ledger.
+- Claims: Complete only across the explicitly listed modules and evidence layers. “Confirmed” requires independent supporting evidence; production/device/security claims require matching production/device/security evidence.
@@ -15,7 +15,7 @@ Before any substantive output, print these exact labels in this exact order. Res
 ```text
 Target: <concrete repository, product, URL, device, PR, or artifact>
 Mode: <selected top-level mode>
-Depth: <quick, standard, or deep>
+Depth: <readiness, standard, or deep>
 Mutation: <report-only or exact authorized mutation boundary>
 Active modules: <comma-separated internal specialist modules>
 Skipped modules: <comma-separated non-active mandatory modules with compact reasons>
@@ -27,7 +27,7 @@ Web context: <none, optional, local-browser, or production>
 1. Infer the mode from product stage, surface, requested artifact, mutation authorization, evidence needs, and deployment state. Do not route by keyword alone.
 2. Refine the public mode to the smallest applicable internal specialist set, then print the required execution header before any substantive output.
 3. Read each active module in full from the path shown in the mode/alias tables. Its specialist body, behavioral contract, STOP gates, and appended upstream judgment ports are binding. Read a lazy specialist phase in full only when the workflow reaches its package-local reference.
-4. Read `references/SHARED-JUDGMENT.md` and `references/AUTHORITY-POLICY.md` for every invocation. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
+4. Read `references/EXECUTION-PROFILES.md`, `references/SHARED-JUDGMENT.md`, and `references/AUTHORITY-POLICY.md` for every invocation. Infer Depth from structured operating conditions, then obey its mandatory modules, legal skips, artifacts, and claim limits. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
 5. If an old asset path is unavailable, use `references/ASSETS.md`. If legacy prose invokes another retired skill, resolve it through `references/COMPATIBILITY.md` and stay inside these six dispatchers.
 6. Preserve report-only versus mutation boundaries. Missing mutation authorization fails closed: do not edit merely because a specialist can fix. Commits, pushes, PRs, merges, deploys, messages, and other external mutations still require affirmative authority from the user.
 7. Match the user's language. Keep code identifiers, commands, and source quotations original when translation would reduce accuracy.
@@ -0,0 +1,28 @@
+<!-- GENERATED by scripts/gstack2/generate-skill-tree.ts; do not edit. -->
+# Inferred execution profiles
+
+Choose a profile from product stage, mutation authority, risk/evidence needs, and deployment state. Prompt keywords and a request to “be quick” are not routing evidence. A profile narrows or strengthens evidence; it never overrides a specialist’s binding question order, pressure, gates, mutation boundary, or exit behavior.
+
+## Smoke/readiness
+
+- Infer when: A narrow, reversible, pre-deployment or operational readiness decision needs bounded evidence; the requested claim is readiness, not completeness.
+- Mandatory modules: Every selected specialist module remains mandatory. Run its entry gates and the smallest source-authorized evidence path that can answer the readiness question.
+- Legal skips: Only source-declared conditional work whose condition is demonstrably false, or evidence unavailable after a named attempt. Never skip a STOP gate, approval boundary, reproduction/root-cause gate, or required physical-device/production evidence.
+- Artifacts: A readiness record naming the exact scope, probes run, evidence and freshness, failures, skipped work with reasons, and the next standard/deep step.
+- Claims: Only ready/not-ready for the named bounded decision. Must say “Readiness profile — not a complete review.” Never claim comprehensive, fully verified, production-safe, or no issues found outside the inspected scope.
+
+## Standard
+
+- Infer when: Normal feature or change work has bounded scope and risk and needs the selected specialist’s complete default workflow.
+- Mandatory modules: Every selected specialist module and all of its mandatory phases, gates, artifacts, and exit checks.
+- Legal skips: Only smart skips explicitly authorized by the specialist and supported by inspected evidence; list each skipped primary module and each skipped conditional phase.
+- Artifacts: Every artifact required by the selected specialist, plus evidence provenance, unresolved decisions, and a skip ledger.
+- Claims: Complete only for the named specialist scope and evidence layer. Broader product, security, production, or device claims require those modules and evidence.
+
+## Deep
+
+- Infer when: Risk, ambiguity, blast radius, cross-system effects, irreversible mutation, security/reliability needs, or production deployment demands stronger evidence.
+- Mandatory modules: Every selected specialist module, all mandatory phases and outside-voice/cross-consumer modules selected by the dispatcher, with unchanged STOP and approval gates.
+- Legal skips: Only specialist-authorized smart skips proven irrelevant. Missing, stale, malformed, or contradictory evidence is a gap or blocker, never a successful skip.
+- Artifacts: All specialist artifacts plus changed-input/unchanged-consumer trace, negative and failure-path evidence, provenance/freshness, rollback or reversibility evidence where applicable, and unresolved-risk ledger.
+- Claims: Complete only across the explicitly listed modules and evidence layers. “Confirmed” requires independent supporting evidence; production/device/security claims require matching production/device/security evidence.
@@ -15,7 +15,7 @@ Before any substantive output, print these exact labels in this exact order. Res
 ```text
 Target: <concrete repository, product, URL, device, PR, or artifact>
 Mode: <selected top-level mode>
-Depth: <quick, standard, or deep>
+Depth: <readiness, standard, or deep>
 Mutation: <report-only or exact authorized mutation boundary>
 Active modules: <comma-separated internal specialist modules>
 Skipped modules: <comma-separated non-active mandatory modules with compact reasons>
@@ -27,7 +27,7 @@ Web context: <none, optional, local-browser, or production>
 1. Infer the mode from product stage, surface, requested artifact, mutation authorization, evidence needs, and deployment state. Do not route by keyword alone.
 2. Refine the public mode to the smallest applicable internal specialist set, then print the required execution header before any substantive output.
 3. Read each active module in full from the path shown in the mode/alias tables. Its specialist body, behavioral contract, STOP gates, and appended upstream judgment ports are binding. Read a lazy specialist phase in full only when the workflow reaches its package-local reference.
-4. Read `references/SHARED-JUDGMENT.md` and `references/AUTHORITY-POLICY.md` for every invocation. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
+4. Read `references/EXECUTION-PROFILES.md`, `references/SHARED-JUDGMENT.md`, and `references/AUTHORITY-POLICY.md` for every invocation. Infer Depth from structured operating conditions, then obey its mandatory modules, legal skips, artifacts, and claim limits. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
 5. If an old asset path is unavailable, use `references/ASSETS.md`. If legacy prose invokes another retired skill, resolve it through `references/COMPATIBILITY.md` and stay inside these six dispatchers.
 6. Preserve report-only versus mutation boundaries. Missing mutation authorization fails closed: do not edit merely because a specialist can fix. Commits, pushes, PRs, merges, deploys, messages, and other external mutations still require affirmative authority from the user.
 7. Match the user's language. Keep code identifiers, commands, and source quotations original when translation would reduce accuracy.
@@ -0,0 +1,28 @@
+<!-- GENERATED by scripts/gstack2/generate-skill-tree.ts; do not edit. -->
+# Inferred execution profiles
+
+Choose a profile from product stage, mutation authority, risk/evidence needs, and deployment state. Prompt keywords and a request to “be quick” are not routing evidence. A profile narrows or strengthens evidence; it never overrides a specialist’s binding question order, pressure, gates, mutation boundary, or exit behavior.
+
+## Smoke/readiness
+
+- Infer when: A narrow, reversible, pre-deployment or operational readiness decision needs bounded evidence; the requested claim is readiness, not completeness.
+- Mandatory modules: Every selected specialist module remains mandatory. Run its entry gates and the smallest source-authorized evidence path that can answer the readiness question.
+- Legal skips: Only source-declared conditional work whose condition is demonstrably false, or evidence unavailable after a named attempt. Never skip a STOP gate, approval boundary, reproduction/root-cause gate, or required physical-device/production evidence.
+- Artifacts: A readiness record naming the exact scope, probes run, evidence and freshness, failures, skipped work with reasons, and the next standard/deep step.
+- Claims: Only ready/not-ready for the named bounded decision. Must say “Readiness profile — not a complete review.” Never claim comprehensive, fully verified, production-safe, or no issues found outside the inspected scope.
+
+## Standard
+
+- Infer when: Normal feature or change work has bounded scope and risk and needs the selected specialist’s complete default workflow.
+- Mandatory modules: Every selected specialist module and all of its mandatory phases, gates, artifacts, and exit checks.
+- Legal skips: Only smart skips explicitly authorized by the specialist and supported by inspected evidence; list each skipped primary module and each skipped conditional phase.
+- Artifacts: Every artifact required by the selected specialist, plus evidence provenance, unresolved decisions, and a skip ledger.
+- Claims: Complete only for the named specialist scope and evidence layer. Broader product, security, production, or device claims require those modules and evidence.
+
+## Deep
+
+- Infer when: Risk, ambiguity, blast radius, cross-system effects, irreversible mutation, security/reliability needs, or production deployment demands stronger evidence.
+- Mandatory modules: Every selected specialist module, all mandatory phases and outside-voice/cross-consumer modules selected by the dispatcher, with unchanged STOP and approval gates.
+- Legal skips: Only specialist-authorized smart skips proven irrelevant. Missing, stale, malformed, or contradictory evidence is a gap or blocker, never a successful skip.
+- Artifacts: All specialist artifacts plus changed-input/unchanged-consumer trace, negative and failure-path evidence, provenance/freshness, rollback or reversibility evidence where applicable, and unresolved-risk ledger.
+- Claims: Complete only across the explicitly listed modules and evidence layers. “Confirmed” requires independent supporting evidence; production/device/security claims require matching production/device/security evidence.
@@ -15,7 +15,7 @@ Before any substantive output, print these exact labels in this exact order. Res
 ```text
 Target: <concrete repository, product, URL, device, PR, or artifact>
 Mode: <selected top-level mode>
-Depth: <quick, standard, or deep>
+Depth: <readiness, standard, or deep>
 Mutation: <report-only or exact authorized mutation boundary>
 Active modules: <comma-separated internal specialist modules>
 Skipped modules: <comma-separated non-active mandatory modules with compact reasons>
@@ -27,7 +27,7 @@ Web context: <none, optional, local-browser, or production>
 1. Infer the mode from product stage, surface, requested artifact, mutation authorization, evidence needs, and deployment state. Do not route by keyword alone.
 2. Refine the public mode to the smallest applicable internal specialist set, then print the required execution header before any substantive output.
 3. Read each active module in full from the path shown in the mode/alias tables. Its specialist body, behavioral contract, STOP gates, and appended upstream judgment ports are binding. Read a lazy specialist phase in full only when the workflow reaches its package-local reference.
-4. Read `references/SHARED-JUDGMENT.md` and `references/AUTHORITY-POLICY.md` for every invocation. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
+4. Read `references/EXECUTION-PROFILES.md`, `references/SHARED-JUDGMENT.md`, and `references/AUTHORITY-POLICY.md` for every invocation. Infer Depth from structured operating conditions, then obey its mandatory modules, legal skips, artifacts, and claim limits. Read `references/RUNTIME.md` before capability-dependent work and `references/WEB-CONTEXT.md` before public-web work.
 5. If an old asset path is unavailable, use `references/ASSETS.md`. If legacy prose invokes another retired skill, resolve it through `references/COMPATIBILITY.md` and stay inside these six dispatchers.
 6. Preserve report-only versus mutation boundaries. Missing mutation authorization fails closed: do not edit merely because a specialist can fix. Commits, pushes, PRs, merges, deploys, messages, and other external mutations still require affirmative authority from the user.
 7. Match the user's language. Keep code identifiers, commands, and source quotations original when translation would reduce accuracy.
@@ -0,0 +1,28 @@
+<!-- GENERATED by scripts/gstack2/generate-skill-tree.ts; do not edit. -->
+# Inferred execution profiles
+
+Choose a profile from product stage, mutation authority, risk/evidence needs, and deployment state. Prompt keywords and a request to “be quick” are not routing evidence. A profile narrows or strengthens evidence; it never overrides a specialist’s binding question order, pressure, gates, mutation boundary, or exit behavior.
+
+## Smoke/readiness
+
+- Infer when: A narrow, reversible, pre-deployment or operational readiness decision needs bounded evidence; the requested claim is readiness, not completeness.
+- Mandatory modules: Every selected specialist module remains mandatory. Run its entry gates and the smallest source-authorized evidence path that can answer the readiness question.
+- Legal skips: Only source-declared conditional work whose condition is demonstrably false, or evidence unavailable after a named attempt. Never skip a STOP gate, approval boundary, reproduction/root-cause gate, or required physical-device/production evidence.
+- Artifacts: A readiness record naming the exact scope, probes run, evidence and freshness, failures, skipped work with reasons, and the next standard/deep step.
+- Claims: Only ready/not-ready for the named bounded decision. Must say “Readiness profile — not a complete review.” Never claim comprehensive, fully verified, production-safe, or no issues found outside the inspected scope.
+
+## Standard
+
+- Infer when: Normal feature or change work has bounded scope and risk and needs the selected specialist’s complete default workflow.
+- Mandatory modules: Every selected specialist module and all of its mandatory phases, gates, artifacts, and exit checks.
+- Legal skips: Only smart skips explicitly authorized by the specialist and supported by inspected evidence; list each skipped primary module and each skipped conditional phase.
+- Artifacts: Every artifact required by the selected specialist, plus evidence provenance, unresolved decisions, and a skip ledger.
+- Claims: Complete only for the named specialist scope and evidence layer. Broader product, security, production, or device claims require those modules and evidence.
+
+## Deep
+
+- Infer when: Risk, ambiguity, blast radius, cross-system effects, irreversible mutation, security/reliability needs, or production deployment demands stronger evidence.
+- Mandatory modules: Every selected specialist module, all mandatory phases and outside-voice/cross-consumer modules selected by the dispatcher, with unchanged STOP and approval gates.
+- Legal skips: Only specialist-authorized smart skips proven irrelevant. Missing, stale, malformed, or contradictory evidence is a gap or blocker, never a successful skip.
+- Artifacts: All specialist artifacts plus changed-input/unchanged-consumer trace, negative and failure-path evidence, provenance/freshness, rollback or reversibility evidence where applicable, and unresolved-risk ledger.
+- Claims: Complete only across the explicitly listed modules and evidence layers. “Confirmed” requires independent supporting evidence; production/device/security claims require matching production/device/security evidence.