@the-bearded-bear/claude-craft 8.26.0 → 8.27.0

package/Dev/i18n/en/Common/templates/settings.json.template

@@ -25,6 +25,10 @@
     ],
     "deny": [
       "Bash(rm -rf:*)",
+      "Bash(dd:*)",
+      "Bash(mkfs:*)",
+      "Bash(git push --force*)",
+      "Bash(git push -f*)",
       "Bash(sudo:*)",
       "Bash(chmod 777:*)",
       "Bash(curl | sh:*)",

package/Dev/i18n/en/Python/skills/coding-standards/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: coding-standards
-description: Rule 03: Coding Standards. Use when reviewing code style or formatting.
+description: "Rule 03: Coding Standards. Use when reviewing code style or formatting."
 ---
 
 # Rule 03: Coding Standards

package/Dev/i18n/en/Python/skills/testing-python/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: testing-python
-description: Rule 07: Testing. Use when writing tests, reviewing test coverage, or setting up testing.
+description: "Rule 07: Testing. Use when writing tests, reviewing test coverage, or setting up testing."
 ---
 
 # Rule 07: Testing

package/Dev/i18n/en/Python/skills/tooling/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: tooling
-description: Rule 06: Tooling. Use when working with tooling.
+description: "Rule 06: Tooling. Use when working with tooling."
 ---
 
 # Rule 06: Tooling

package/Dev/i18n/en/Symfony/commands/check-security.md

@@ -55,7 +55,7 @@ docker run --rm -v $(pwd):/app php:8.5-cli cat /app/config/packages/security.yam
 
 ### Step 3 : OWASP Top 10 - Injection
 
-#### A03:2021 – Injection (SQL, NoSQL, OS, LDAP) (3 points)
+#### A05:2025 – Injection (SQL, NoSQL, OS, LDAP) (3 points)
 
 ```bash
 # Vérifier l'utilisation de requêtes préparées
@@ -79,9 +79,9 @@ docker run --rm -v $(pwd):/app php:8.5-cli grep -rE "\"SELECT.*\\..*\$" /app/src
 
 **Points obtained** : ___/3
 
-### Step 4 : OWASP Top 10 - Broken Authentication
+### Step 4 : OWASP Top 10 - Authentication Failures
 
-#### A07:2021 – Identification and Authentication Failures (3 points)
+#### A07:2025 – Authentication Failures (3 points)
 
 ```bash
 # Vérifier la configuration des mots de passe
@@ -104,9 +104,9 @@ docker run --rm -v $(pwd):/app php:8.5-cli grep -r "RateLimiter" /app/config --i
 
 **Points obtained** : ___/3
 
-### Step 5 : OWASP Top 10 - Sensitive Data Exposure
+### Step 5 : OWASP Top 10 - Cryptographic Failures
 
-#### A02:2021 – Cryptographic Failures (3 points)
+#### A04:2025 – Cryptographic Failures (3 points)
 
 ```bash
 # Vérifier les secrets dans le code
@@ -131,7 +131,7 @@ docker run --rm -v $(pwd):/app php:8.5-cli grep -r "SECURE_SCHEME" /app/.env.exa
 
 ### Step 6 : OWASP Top 10 - Broken Access Control
 
-#### A01:2021 – Broken Access Control (3 points)
+#### A01:2025 – Broken Access Control (3 points)
 
 ```bash
 # Vérifier les Voters
@@ -156,7 +156,7 @@ docker run --rm -v $(pwd):/app php:8.5-cli grep -r "@IsGranted" /app/src --inclu
 
 ### Step 7 : OWASP Top 10 - XSS et CSRF
 
-#### A03:2021 – XSS (Cross-Site Scripting) (2 points)
+#### A05:2025 – Injection : XSS (Cross-Site Scripting) (2 points)
 
 ```bash
 # Vérifier l'auto-échappement Twig
@@ -179,7 +179,7 @@ docker run --rm -v $(pwd):/app php:8.5-cli grep -r "|raw" /app/templates --inclu
 
 **Points obtained** : ___/2
 
-#### A08:2021 – CSRF (Cross-Site Request Forgery) (2 points)
+#### A01:2025 – Broken Access Control : CSRF (Cross-Site Request Forgery) (2 points)
 
 ```bash
 # Vérifier CSRF protection
@@ -201,7 +201,7 @@ docker run --rm -v $(pwd):/app php:8.5-cli grep -r "csrf_protection" /app/config
 
 ### Step 8 : OWASP Top 10 - Autres Vulnérabilités
 
-#### A05:2021 – Security Misconfiguration (2 points)
+#### A02:2025 – Security Misconfiguration (2 points)
 
 ```bash
 # Vérifier le mode debug
@@ -224,7 +224,7 @@ docker run --rm -v $(pwd):/app php:8.5-cli composer audit
 
 **Points obtained** : ___/2
 
-#### A06:2021 – Vulnerable and Outdated Components (1 point)
+#### A03:2025 – Software Supply Chain Failures (1 point)
 
 ```bash
 # Audit de sécurité Composer
@@ -398,16 +398,16 @@ Fixed in: 5.4.25
    OWASP TOP 10 - RÉSUMÉ
 =================================================
 
-A01:2021 - Broken Access Control          : [✅|⚠️|❌]
-A02:2021 - Cryptographic Failures         : [✅|⚠️|❌]
-A03:2021 - Injection                      : [✅|⚠️|❌]
-A04:2021 - Insecure Design                : [✅|⚠️|❌]
-A05:2021 - Security Misconfiguration      : [✅|⚠️|❌]
-A06:2021 - Vulnerable Components          : [✅|⚠️|❌]
-A07:2021 - Authentication Failures        : [✅|⚠️|❌]
-A08:2021 - Software and Data Integrity    : [✅|⚠️|❌]
-A09:2021 - Security Logging Failures      : [✅|⚠️|❌]
-A10:2021 - Server-Side Request Forgery    : [✅|⚠️|❌]
+A01:2025 - Broken Access Control                 : [✅|⚠️|❌]
+A02:2025 - Security Misconfiguration             : [✅|⚠️|❌]
+A03:2025 - Software Supply Chain Failures        : [✅|⚠️|❌]
+A04:2025 - Cryptographic Failures                : [✅|⚠️|❌]
+A05:2025 - Injection                             : [✅|⚠️|❌]
+A06:2025 - Insecure Design                       : [✅|⚠️|❌]
+A07:2025 - Authentication Failures               : [✅|⚠️|❌]
+A08:2025 - Software or Data Integrity Failures   : [✅|⚠️|❌]
+A09:2025 - Logging & Alerting Failures           : [✅|⚠️|❌]
+A10:2025 - Mishandling of Exceptional Conditions : [✅|⚠️|❌]
 
 =================================================
    CONFORMITÉ RGPD

package/Dev/i18n/en/Symfony/skills/aggregates/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: aggregates
-description: Règle 05 : Aggregates et Aggregate Roots. Use when implementing DDD patterns.
+description: "Règle 05 : Aggregates et Aggregate Roots. Use when implementing DDD patterns."
 ---
 
 # Règle 05 : Aggregates et Aggregate Roots

package/Dev/i18n/en/Symfony/skills/async/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: async
-description: Règle 10 : Async-First avec Symfony Messenger. Use when working with async.
+description: "Règle 10 : Async-First avec Symfony Messenger. Use when working with async."
 ---
 
 # Règle 10 : Async-First avec Symfony Messenger

package/Dev/i18n/en/Symfony/skills/cqrs/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: cqrs
-description: Règle 07 : CQRS - Command Query Responsibility Segregation. Use when implementing DDD patterns.
+description: "Règle 07 : CQRS - Command Query Responsibility Segregation. Use when implementing DDD patterns."
 ---
 
 # Règle 07 : CQRS - Command Query Responsibility Segregation

package/Dev/i18n/en/Symfony/skills/domain-events/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: domain-events
-description: Règle 06 : Domain Events. Use when implementing DDD patterns.
+description: "Règle 06 : Domain Events. Use when implementing DDD patterns."
 ---
 
 # Règle 06 : Domain Events

package/Dev/i18n/en/Symfony/skills/i18n/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: i18n
-description: Règle 09 : Internationalisation (i18n). Use when implementing internationalization.
+description: "Règle 09 : Internationalisation (i18n). Use when implementing internationalization."
 ---
 
 # Règle 09 : Internationalisation (i18n)

package/Dev/i18n/en/Symfony/skills/multitenant/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: multitenant
-description: Règle 08 : Multitenant - Isolation et Sécurité. Use when working with multitenant.
+description: "Règle 08 : Multitenant - Isolation et Sécurité. Use when working with multitenant."
 ---
 
 # Règle 08 : Multitenant - Isolation et Sécurité

package/Dev/i18n/en/Symfony/skills/value-objects/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: value-objects
-description: Règle 04 : Value Objects. Use when implementing DDD patterns.
+description: "Règle 04 : Value Objects. Use when implementing DDD patterns."
 ---
 
 # Règle 04 : Value Objects

package/Dev/i18n/en/Team/commands/audit.md

@@ -1,6 +1,6 @@
 ---
-description: Full Audit Team - Parallel multi-technology audit using Agent Teams
-argument-hint: "[--techs=auto|tech1,tech2] [--max-workers=4]"
+description: Full Audit Team - Parallel multi-technology audit using sub-agents (Agent Teams optional)
+argument-hint: "[--techs=auto|tech1,tech2] [--max-workers=4] [--engine=subagents|teams]"
 ---
 
 # Full Audit Team - Parallel Multi-Technology Audit
@@ -13,16 +13,41 @@ $ARGUMENTS
 
 - `--techs=auto`: Auto-detect technologies (default). Or specify comma-separated: `--techs=symfony,react`
 - `--max-workers=4`: Maximum parallel auditor workers (default: 4, max: 4)
+- `--engine=subagents`: Execution engine (default: `subagents`). Options: `subagents` (delegation to sub-agents via the `Agent` tool, no prerequisite), `teams` (experimental Agent Teams). See "Execution engine"
 - `--output-dir=<path>`: Custom output directory for audit results
 - `--max-cost=<dollars>`: Maximum budget in dollars. If the estimated parallel cost exceeds this threshold, execution is blocked with an OVER BUDGET message
 - `--dry-run`: Show team composition and estimated cost without executing
 - `--skip-aggregation`: Output per-stack results without merging
 - `--sequential`: Run audits sequentially instead of in parallel (no Agent Teams overhead). Useful for single-technology projects or when Agent Teams is not available.
 
+## Execution engine (`--engine`)
+
+The default engine relies on no experimental feature: the command works even when `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is not set.
+
+| Engine | Activation | Prerequisites | Behavior |
+|--------|------------|---------------|----------|
+| `subagents` (default) | no flag or `--engine=subagents` | None (native `Agent` tool) | The current session acts as the lead and delegates each worker task to a sub-agent via the `Agent` tool (worker model passed through the `model` parameter). Independent sub-agents are launched in a single message so they run in parallel; each one writes its result to its isolated directory and returns only a short summary. Workers do not talk to each other: the lead relays. Works interactively, with `claude -p` and in CI. |
+| `teams` | `--engine=teams` | `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`, Claude Code v2.1.32+ | Agent Teams team: persistent workers coordinated through `TaskCreate` / `TaskList` / `SendMessage`, direct teammate messaging, `TeammateIdle` / `TaskCompleted` hooks. |
+
+**Fallback**: if `--engine=teams` is requested while `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is missing, do not fail — print `ℹ️ Agent Teams unavailable: falling back to --engine=subagents` and continue with the `subagents` engine.
+
+**`teams` engine limits**: no non-interactive mode (`claude -p`) and no CI runs; one team per session (finish the current team before starting another); research preview API, subject to change.
+
+The process below is written with Agent Teams primitives. With `--engine=subagents`, apply this mapping:
+
+| Agent Teams (`teams`) | Sub-agents (`subagents`) |
+|------------------------|---------------------------|
+| `TaskCreate` + worker spawn | One `Agent` tool call whose prompt carries the task description |
+| `TaskList` polling / synchronization barrier | Wait for every `Agent` call of the batch to return |
+| `SendMessage` to a worker | Resume the sub-agent (`SendMessage` to its ID) or a new `Agent` call with the needed context |
+| `shutdown_request` | Not needed: a sub-agent ends when it returns |
+
+**Audit specifics**: one sub-agent per stack, using the spawn template prompt from Step 4. `--sequential` remains available to process everything in the current session, without delegation.
+
 ## Prerequisites
 
-- Claude Code v2.1.32+ with Agent Teams support
-- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set
+- Claude Code v2.1.32+ with Agent Teams support — only with `--engine=teams`
+- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set — only with `--engine=teams`
 - Project with 2+ detected technology stacks (single-stack projects should use `--sequential` flag)
 - `Tools/AgentTeams/lib/compatibility-check.sh` available
 - `Tools/AgentTeams/lib/result-aggregator.sh` available

package/Dev/i18n/en/Team/commands/delivery.md

@@ -1,6 +1,6 @@
 ---
-description: Delivery Team - Full sprint lifecycle (writing + implementation) using Agent Teams
-argument-hint: <sprint-name|prd-path> [--phase=all|writing|implementation] [--max-workers=3]
+description: Delivery Team - Full sprint lifecycle (writing + implementation) using sub-agents (Agent Teams optional)
+argument-hint: <sprint-name|prd-path> [--phase=all|writing|implementation] [--max-workers=3] [--engine=subagents|teams]
 ---
 
 # Delivery Team - Full Sprint Lifecycle (Writing + Implementation)
@@ -14,6 +14,7 @@ $ARGUMENTS
 - `<sprint-name|prd-path>`: Sprint name/ID or path to the PRD document
 - `--phase=all`: Phase to execute (default: `all`). Options: `all`, `writing`, `implementation`
 - `--max-workers=3`: Maximum parallel workers per phase (default: 3, max: 3)
+- `--engine=subagents`: Execution engine (default: `subagents`). Options: `subagents` (delegation to sub-agents via the `Agent` tool, no prerequisite), `teams` (experimental Agent Teams). See "Execution engine"
 - `--overnight`: Run in overnight mode (bounded, stops at 6am)
 - `--supervised`: Pause before each story for human confirmation
 - `--max-stories=10`: Maximum stories to process (default: 10)
@@ -23,10 +24,34 @@ $ARGUMENTS
 - `--max-rewrites=2`: Maximum rewrite loops per artefact in Phase 1 (default: 2)
 - `--max-cost=<dollars>`: Maximum budget in dollars. If the estimated parallel cost exceeds this threshold, execution is blocked with an OVER BUDGET message
 
+## Execution engine (`--engine`)
+
+The default engine relies on no experimental feature: the command works even when `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is not set.
+
+| Engine | Activation | Prerequisites | Behavior |
+|--------|------------|---------------|----------|
+| `subagents` (default) | no flag or `--engine=subagents` | None (native `Agent` tool) | The current session acts as the lead and delegates each worker task to a sub-agent via the `Agent` tool (worker model passed through the `model` parameter). Independent sub-agents are launched in a single message so they run in parallel; each one writes its result to its isolated directory and returns only a short summary. Workers do not talk to each other: the lead relays. Works interactively, with `claude -p` and in CI. |
+| `teams` | `--engine=teams` | `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`, Claude Code v2.1.32+ | Agent Teams team: persistent workers coordinated through `TaskCreate` / `TaskList` / `SendMessage`, direct teammate messaging, `TeammateIdle` / `TaskCompleted` hooks. |
+
+**Fallback**: if `--engine=teams` is requested while `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is missing, do not fail — print `ℹ️ Agent Teams unavailable: falling back to --engine=subagents` and continue with the `subagents` engine.
+
+**`teams` engine limits**: no non-interactive mode (`claude -p`) and no CI runs; one team per session (finish the current team before starting another); research preview API, subject to change.
+
+The process below is written with Agent Teams primitives. With `--engine=subagents`, apply this mapping:
+
+| Agent Teams (`teams`) | Sub-agents (`subagents`) |
+|------------------------|---------------------------|
+| `TaskCreate` + worker spawn | One `Agent` tool call whose prompt carries the task description |
+| `TaskList` polling / synchronization barrier | Wait for every `Agent` call of the batch to return |
+| `SendMessage` to a worker | Resume the sub-agent (`SendMessage` to its ID) or a new `Agent` call with the needed context |
+| `shutdown_request` | Not needed: a sub-agent ends when it returns |
+
+**Delivery specifics**: in Phase 1, cross-review goes through the lead, which hands the drafted artifact to a reviewer sub-agent and sends its feedback back to the writer. In Phase 2, launch each developer worker with `isolation: "worktree"`.
+
 ## Prerequisites
 
-- Claude Code v2.1.32+ with Agent Teams support
-- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set
+- Claude Code v2.1.32+ with Agent Teams support — only with `--engine=teams`
+- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set — only with `--engine=teams`
 - PRD or tech spec available (for Phase 1) or BMAD sprint backlog with `ready-for-dev` stories (for Phase 2 only)
 - Sprint metadata in `.bmad/sprint-status.yaml`
 - `Tools/AgentTeams/lib/compatibility-check.sh` available

package/Dev/i18n/en/Team/commands/security.md

@@ -1,6 +1,6 @@
 ---
-description: "Security Review Team - Parallel multi-dimension security audit using Agent Teams"
-argument-hint: "[--scope=full|code|deps|infra] [--max-workers=3]"
+description: "Security Review Team - Parallel multi-dimension security audit using sub-agents (Agent Teams optional)"
+argument-hint: "[--scope=full|code|deps|infra] [--max-workers=3] [--engine=subagents|teams]"
 ---
 
 # Security Review Team - Parallel Multi-Dimension Security Audit
@@ -13,16 +13,43 @@ $ARGUMENTS
 
 - `--scope=full`: Audit scope (default: `full`). Options: `full`, `code`, `deps`, `infra`
 - `--max-workers=3`: Maximum parallel reviewers (default: 3, max: 3)
+- `--engine=subagents`: Execution engine (default: `subagents`). Options: `subagents` (delegation to sub-agents via the `Agent` tool, no prerequisite), `teams` (experimental Agent Teams). See "Execution engine"
 - `--severity=medium`: Minimum severity to report: `low`, `medium`, `high`, `critical`
 - `--output-dir=<path>`: Custom output directory for security results
 - `--dry-run`: Show team composition and scan plan without executing
 - `--sarif`: Output results in SARIF format (for CI/CD integration)
 - `--max-cost=<dollars>`: Maximum budget in dollars. If the estimated parallel cost exceeds this threshold, execution is blocked with an OVER BUDGET message
 
+## Execution engine (`--engine`)
+
+The default engine relies on no experimental feature: the command works even when `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is not set.
+
+| Engine | Activation | Prerequisites | Behavior |
+|--------|------------|---------------|----------|
+| `subagents` (default) | no flag or `--engine=subagents` | None (native `Agent` tool) | The current session acts as the lead and delegates each worker task to a sub-agent via the `Agent` tool (worker model passed through the `model` parameter). Independent sub-agents are launched in a single message so they run in parallel; each one writes its result to its isolated directory and returns only a short summary. Workers do not talk to each other: the lead relays. Works interactively, with `claude -p` and in CI. |
+| `teams` | `--engine=teams` | `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`, Claude Code v2.1.32+ | Agent Teams team: persistent workers coordinated through `TaskCreate` / `TaskList` / `SendMessage`, direct teammate messaging, `TeammateIdle` / `TaskCompleted` hooks. |
+
+**Fallback**: if `--engine=teams` is requested while `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is missing, do not fail — print `ℹ️ Agent Teams unavailable: falling back to --engine=subagents` and continue with the `subagents` engine.
+
+**`teams` engine limits**: no non-interactive mode (`claude -p`) and no CI runs; one team per session (finish the current team before starting another); research preview API, subject to change.
+
+The process below is written with Agent Teams primitives. With `--engine=subagents`, apply this mapping:
+
+| Agent Teams (`teams`) | Sub-agents (`subagents`) |
+|------------------------|---------------------------|
+| `TaskCreate` + worker spawn | One `Agent` tool call whose prompt carries the task description |
+| `TaskList` polling / synchronization barrier | Wait for every `Agent` call of the batch to return |
+| `SendMessage` to a worker | Resume the sub-agent (`SendMessage` to its ID) or a new `Agent` call with the needed context |
+| `shutdown_request` | Not needed: a sub-agent ends when it returns |
+
+**Scope**: `--scope=full|code|deps|infra` applies to both engines (`full` = 3 reviewers in parallel, a single scope = a single reviewer).
+
+**Difference with native `/security-review`**: Claude Code's built-in `/security-review` only analyzes the diff (pending changes on the current branch). `/team:security` audits the whole repository for the chosen scope (code, dependencies, infrastructure). Use `/security-review` before a commit or a PR, `/team:security` for a full audit.
+
 ## Prerequisites
 
-- Claude Code v2.1.32+ with Agent Teams support
-- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set
+- Claude Code v2.1.32+ with Agent Teams support — only with `--engine=teams`
+- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set — only with `--engine=teams`
 - Docker available for running security scanners
 - `Tools/AgentTeams/lib/compatibility-check.sh` available
 - `Tools/AgentTeams/lib/result-aggregator.sh` available

package/Dev/i18n/en/Team/commands/sprint.md

@@ -1,6 +1,6 @@
 ---
-description: Sprint Development Team - Parallel story implementation using Agent Teams
-argument-hint: <sprint-name> [--max-workers=3] [--overnight]
+description: Sprint Development Team - Parallel story implementation using sub-agents (Agent Teams optional)
+argument-hint: <sprint-name> [--max-workers=3] [--overnight] [--engine=subagents|teams]
 ---
 
 # Sprint Development Team - Parallel Story Implementation
@@ -13,18 +13,43 @@ $ARGUMENTS
 
 - `<sprint-name>`: Name or ID of the sprint to process
 - `--max-workers=3`: Maximum parallel dev workers (default: 2, max: 3)
+- `--engine=subagents`: Execution engine (default: `subagents`). Options: `subagents` (delegation to sub-agents via the `Agent` tool, no prerequisite), `teams` (experimental Agent Teams). See "Execution engine"
 - `--overnight`: Run in overnight mode (bounded, stops at 6am)
 - `--supervised`: Pause before each story for human confirmation
 - `--max-stories=10`: Maximum stories to process (default: 10)
 - `--timeout=12`: Maximum runtime in hours (default: 12)
 - `--dry-run`: Show team composition and story assignments without executing
 - `--max-cost=<dollars>`: Maximum budget in dollars. If the estimated parallel cost exceeds this threshold, execution is blocked with an OVER BUDGET message
-- `--ralph-mode`: Enable Ralph recovery engine (error classification, auto-retry, escalation service) alongside Agent Teams parallelism.
+- `--ralph-mode`: Enable Ralph recovery engine (error classification, auto-retry, escalation service) alongside Agent Teams parallelism. Requires `--engine=teams`: the Ralph Teams adapter relies on `TaskCreate` / `SendMessage`.
+
+## Execution engine (`--engine`)
+
+The default engine relies on no experimental feature: the command works even when `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is not set.
+
+| Engine | Activation | Prerequisites | Behavior |
+|--------|------------|---------------|----------|
+| `subagents` (default) | no flag or `--engine=subagents` | None (native `Agent` tool) | The current session acts as the lead and delegates each worker task to a sub-agent via the `Agent` tool (worker model passed through the `model` parameter). Independent sub-agents are launched in a single message so they run in parallel; each one writes its result to its isolated directory and returns only a short summary. Workers do not talk to each other: the lead relays. Works interactively, with `claude -p` and in CI. |
+| `teams` | `--engine=teams` | `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`, Claude Code v2.1.32+ | Agent Teams team: persistent workers coordinated through `TaskCreate` / `TaskList` / `SendMessage`, direct teammate messaging, `TeammateIdle` / `TaskCompleted` hooks. |
+
+**Fallback**: if `--engine=teams` is requested while `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is missing, do not fail — print `ℹ️ Agent Teams unavailable: falling back to --engine=subagents` and continue with the `subagents` engine.
+
+**`teams` engine limits**: no non-interactive mode (`claude -p`) and no CI runs; one team per session (finish the current team before starting another); research preview API, subject to change.
+
+The process below is written with Agent Teams primitives. With `--engine=subagents`, apply this mapping:
+
+| Agent Teams (`teams`) | Sub-agents (`subagents`) |
+|------------------------|---------------------------|
+| `TaskCreate` + worker spawn | One `Agent` tool call whose prompt carries the task description |
+| `TaskList` polling / synchronization barrier | Wait for every `Agent` call of the batch to return |
+| `SendMessage` to a worker | Resume the sub-agent (`SendMessage` to its ID) or a new `Agent` call with the needed context |
+| `shutdown_request` | Not needed: a sub-agent ends when it returns |
+
+**Sprint specifics**: with `subagents`, launch each developer worker with `isolation: "worktree"` so parallel stories do not edit the same working tree.
 
 ## Prerequisites
 
-- Claude Code v2.1.32+ with Agent Teams support
-- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set
+- Claude Code v2.1.32+ with Agent Teams support — only with `--engine=teams`
+- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` environment variable set — only with `--engine=teams`
 - BMAD sprint backlog with stories in `ready-for-dev` status
 - Sprint metadata in `.bmad/sprint-status.yaml`
 - At least 2 independent stories (single-story sprints use sequential Ralph)

package/Dev/i18n/en/UIUX/commands/audit.md

@@ -24,36 +24,7 @@ Example: `/uiux:audit src/pages/Dashboard.tsx AAA`
 
 ### Step 1: Accessibility Audit (A11y Expert)
 
-#### 1.1 Automated audit
-```bash
-# Run if available
-npx axe-cli {URL}
-npx pa11y {URL}
-# Or check Lighthouse
-```
-
-#### 1.2 Manual WCAG 2.2 AAA verification
-
-**Perceivable**
-- [ ] Images with alt text
-- [ ] Semantic structure (h1-h6, landmarks)
-- [ ] Contrast ≥ 7:1 (AAA)
-- [ ] Reflow at 320px
-
-**Operable**
-- [ ] Complete keyboard navigation
-- [ ] No keyboard trap
-- [ ] Visible focus (≥ 2px)
-- [ ] Touch targets ≥ 44px
-
-**Understandable**
-- [ ] lang on html
-- [ ] Labels on inputs
-- [ ] Clear error messages
-
-**Robust**
-- [ ] Correct ARIA
-- [ ] aria-live for dynamic
+Run `/uiux:a11y-audit {URL or path} {level}` and carry its report (score, failing WCAG criteria, remediations) into the synthesis. Do not repeat the automated audit or the WCAG checklist here: `/uiux:a11y-audit` is the single source.
 
 ### Step 2: UX/Ergonomics Audit (UX Expert)
 

package/Dev/i18n/en/UIUX/commands/component-spec.md

@@ -41,11 +41,7 @@ Define visuals:
 
 ### Step 3: A11y Specification (A11y Expert)
 
-Define accessibility:
-- HTML semantics
-- ARIA attributes
-- Keyboard navigation
-- Screen reader announcements
+Run `/uiux:a11y-component {NAME}` and carry its specification (HTML semantics, ARIA attributes, keyboard navigation, focus, contrast, screen reader announcements) into the ♿ section of the synthesis. Do not redefine it here.
 
 ### Step 4: Synthesis
 

package/Dev/i18n/en/UIUX/commands/orchestrator.md

@@ -1,102 +1,9 @@
 ---
-description: UI/UX Orchestrator
-argument-hint: [arguments]
+description: "Deprecated (removed in 9.0): use the @uiux-orchestrator agent"
+deprecated: true
+replaced-by: "@uiux-orchestrator"
 ---
 
-# UI/UX Orchestrator
+> ⚠️ `/uiux:orchestrator` is deprecated and will be removed in 9.0. Use the `@uiux-orchestrator` agent.
 
-You are the UI/UX Orchestrator. You must coordinate the 3 experts to deliver exceptional interfaces.
-
-## Arguments
-$ARGUMENTS
-
-Arguments:
-- Request type: component, audit, flow, tokens
-- Objective or description
-
-Example: `/uiux:orchestrator component "Date picker"` or `/uiux:orchestrator audit "Checkout page"`
-
-## Plan Mode
-
-> Plan mode is activated automatically when the scope spans multiple modules or requires cross-cutting investigation.
-
-## MISSION
-
-### Step 1: Analyze the request
-
-Identify:
-- Expected deliverable type
-- Expert(s) to involve
-- Order of intervention
-
-### Step 2: Delegate to experts
-
-| Type | Experts | Order |
-|------|---------|-------|
-| New component | UI → UX → A11y | Sequential |
-| Audit | A11y → UX → UI | Sequential |
-| User flow | UX → UI → A11y | Sequential |
-| Design tokens | UI only | Direct |
-
-### Step 3: Consolidate and arbitrate
-
-In case of conflict, apply priority rules:
-1. AAA Accessibility (non-negotiable)
-2. Lighthouse 100/100
-3. UX > Aesthetics
-4. Mobile-first
-5. Design system consistency
-
-### Step 4: Deliver synthesis
-
-```
-══════════════════════════════════════════════════════════════
-📋 UI/UX SYNTHESIS: {TOPIC}
-══════════════════════════════════════════════════════════════
-
-Type: {Component | Audit | Flow | Tokens}
-Date: {date}
-
-──────────────────────────────────────────────────────────────
-🧠 UX
-──────────────────────────────────────────────────────────────
-
-{UX contributions summary}
-
-──────────────────────────────────────────────────────────────
-🎨 UI
-──────────────────────────────────────────────────────────────
-
-{UI contributions summary}
-
-──────────────────────────────────────────────────────────────
-♿ ACCESSIBILITY
-──────────────────────────────────────────────────────────────
-
-{A11y contributions summary}
-
-──────────────────────────────────────────────────────────────
-⚖️ ARBITRATIONS
-──────────────────────────────────────────────────────────────
-
-| Conflict | Decision | Justification |
-|----------|----------|---------------|
-| {conflict} | {decision} | {rule applied} |
-
-──────────────────────────────────────────────────────────────
-✅ VALIDATION CHECKLIST
-──────────────────────────────────────────────────────────────
-
-- [ ] WCAG 2.2 AAA compliant
-- [ ] Lighthouse 100/100 preserved
-- [ ] Mobile-first respected
-- [ ] Only tokens (no hardcode)
-- [ ] All 3 experts consulted
-
-──────────────────────────────────────────────────────────────
-🎯 NEXT STEPS
-──────────────────────────────────────────────────────────────
-
-1. {priority action}
-2. {next action}
-```
+Delegate "$ARGUMENTS" to the `uiux-orchestrator` agent (Agent tool) and return its answer.

package/Dev/i18n/en/Workflow/commands/auto-sprint.md

@@ -1,6 +1,6 @@
 ---
 description: End-to-end sprint orchestrator (start -> decompose -> validate -> implement -> PR -> CI -> review -> retro -> merge)
-argument-hint: "<N> [--from-prd] [--auto-merge] [--max-fix-attempts=2] [--max-workers=3] [--base=main] [--dry-run] [--overnight]"
+argument-hint: "<N> [--from-prd] [--auto-merge] [--max-fix-attempts=2] [--max-workers=3] [--base=main] [--dry-run] [--overnight] [--engine=subagents|teams]"
 ---
 
 # Auto Sprint — End-to-End Sprint Orchestrator
@@ -31,14 +31,15 @@ $ARGUMENTS
   rule 09, and the Karpathy "no auto-merge without human review" principle).
 - `--max-fix-attempts=2` : Max auto-fix retries per failing gate before aborting (default: 2).
 - `--max-workers=3` : Max parallel dev workers in the implementation phase (default: 2, max: 3).
+- `--engine=subagents`: Engine of the implementation phase (default: `subagents`, delegation via the `Agent` tool). `teams`: experimental Agent Teams, falling back to `subagents` when unavailable. See `/team:sprint`, "Execution engine" section
 - `--base=main` : Base branch for the PR (default: `main`).
 - `--dry-run` : Print the planned 9 phases and the resolved sprint context, then stop. **No writes.**
 - `--overnight` : Pass-through to the implementation phase (bounded, stops at 6am).
 
 ## Prerequisites
 
-- Claude Code v2.1.32+ with Agent Teams support
-- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` set
+- Claude Code v2.1.32+ with Agent Teams support — only with `--engine=teams`
+- `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` set — only with `--engine=teams`
 - `gh` CLI authenticated (PR create / checks / merge)
 - Docker available (all tests run via Docker — see project CLAUDE.md)
 - BMAD v6 project: `.bmad/sprint-status.yaml` is created by `/workflow:start` (phase 1) if missing
@@ -106,6 +107,8 @@ Spawn one isolated sub-agent:
 
 Assume the **`/team:sprint` conductor role directly** (do **not** spawn a nested Agent Team):
 
+**Engine**: with `--engine=subagents` (default), in step 4 launch one sub-agent per story via the `Agent` tool (`isolation: "worktree"`, all in a single message) instead of `TaskCreate`, and in step 5 wait for them to return instead of polling `TaskList` (full mapping in `/team:sprint`, "Execution engine" section). With `--engine=teams`, follow the steps as written.
+
 1. Read `.bmad/sprint-status.yaml`; filter stories at `ready-for-dev`.
 2. Analyze file-domain independence (flag `**/Shared/**`, `**/Common/**`, `**/Utils/**`,
    `**/Helpers/**` overlaps → sequence in the same worker).
@@ -195,7 +198,7 @@ Phase            | Status | Notes
 | Story DoD miss after retries | Mark `blocked`, continue, report at the end |
 | CI red after retries | Abort with failing-check report |
 | Merge blocked (protection / scope) | Surface the error + suggested flag, do not force |
-| Agent Teams unavailable | Abort Phase 4 with a setup hint (`CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`) |
+| Agent Teams unavailable with `--engine=teams` | Fall back to `--engine=subagents` (no abort); set `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` to use `--engine=teams` |
 
 ## Notes