aw-spec-cli 0.1.28 → 0.1.30

package/assets/.appweaver/templates/delta-design-template.md

@@ -115,17 +115,19 @@ realisation); new migration needed? (yes/no); any contract change other systems
 ## Contract deltas for the frontend lane
 
 <!--
-  What this change adds to or alters in the service's PUBLISHED surface — the only part
-  of this document the frontend lane may bind against. One row per contract; the frontend
-  delta plan cites these rather than re-deriving them. "None" when the surface is
+  What this change adds to, alters in or removes from the service's PUBLISHED surface — the
+  only part of this document the frontend lane may bind against. One row per contract; the
+  frontend delta plan cites these rather than re-deriving them. "None" when the surface is
   unchanged, which is what a purely internal refactor should say. The operation id is the
   endpoint's name in the published API document, unique across the service; an endpoint
-  the master plan already names keeps that id.
+  the master plan already names keeps that id. A `changed` or `removed` row whose operation
+  the frontend plan binds to is reported by analyze-plan-delta until the frontend delta plan
+  of this change — or a follow-up change — carries the matching change.
 -->
 
 | Operation id | Contract | Change | Shape | Serves |
 |---|---|---|---|---|
-| `[OperationId]` | `[VERB] [/route]` | added / changed | [request → response field shapes] | IC-[nnn] · FR-[nnn] |
+| `[OperationId]` | `[VERB] [/route]` | added / changed / removed | [request → response field shapes] | IC-[nnn] · FR-[nnn] |
 
 ## Other open deltas
 

package/assets/.appweaver/templates/delta-spec-template.md

@@ -8,6 +8,8 @@
 
 **Wave**: [N]   <!-- script-stamped; same wave = developable in parallel, higher wave = after -->
 
+**Lanes**: [backend, frontend]   <!-- script-stamped from the answer specify asks when it opens the change: the lanes that plan and build it. The spec is written whole either way. Changed later with aw-spec-actions cr-lanes --cr-id <id> --set both|backend|frontend -->
+
 **Base version**: [vX.Y]   <!-- the current master version; close-delta bumps the minor at the end of the chain -->
 
 **Model delta**: `specs/[Context]/domain-model.[CR-ID].delta.md`   <!-- "none" when this change is behaviour-only -->

package/assets/.appweaver/templates/design-template.md

@@ -274,6 +274,23 @@ text, and do not add a rule the model does not carry)*
 
 - **Maps to**: `[Command/Query]` (§2.2)
 
+### Requests from the frontend lane
+
+<!-- THE ANSWERS to the frontend plan's §5 "Contract deltas needed" table, one row
+     per RQ-n it raises. PUBLISH: add the endpoint above — under the proposed
+     operation id unless it clashes with one this service already uses — and name
+     the id here. DECLINE: say why in one line; a request the spec does not
+     support is declined as a spec gap for `specify`, never published on a guess.
+     The frontend plan re-binds to the id given here. Keep every answered row;
+     write `None` instead of the table when the frontend plan raises no request.
+     plan and analyze-plan draft the endpoints above, by script, into
+     api-draft.json beside this plan, so a frontend can bind to an endpoint before
+     the service is built. -->
+
+| Request | Outcome | Operation id | Why |
+|---|---|---|---|
+| RQ-[n] | published / declined | `[OperationId]` / — | [one line — required when declined] |
+
 ## 4. Infrastructure Layer
 
 <!--

package/assets/.appweaver/templates/ui-design-template.md

@@ -197,8 +197,10 @@ a lost connection mid-flight.]
      line. An interface the experience needs and the service does not offer is a
      CONTRACT DELTA raised back to the service lane, never a workaround in the
      client. The OPERATION ID is the one backend-plan.md §3 gives the endpoint —
-     the name implement finds it by in the exported API document, and checks
-     before generating. `—` when no service plan names it yet, with a §8 DA-n. -->
+     the name implement finds it by in the API document, and checks before
+     generating. A row that needs an operation the service does not publish yet
+     cites the operation id its request below proposes, and names the RQ-n in
+     its Serves cell. -->
 
 | Screen / component | Operation id | Operation | Sends (field shape) | Receives (field shape) | Failure handling | Serves |
 |---|---|---|---|---|---|---|
@@ -207,8 +209,21 @@ a lost connection mid-flight.]
 **Write semantics**: [per mutating operation — optimistic or pessimistic, what
 it invalidates, and how an optimistic update rolls back.]
 
-**Contract deltas needed**: [anything the experience requires that the service
-surface does not offer, raised back to the service lane. Or `None`.]
+**Contract deltas needed**
+
+<!-- THE REQUEST LEDGER. One row per operation the experience needs that the
+     service's published surface does not offer — and, when there is no backend
+     plan yet, one per binding. Each request keeps its RQ-n for life: never
+     renumber or reuse one. Propose the operation id and the field shapes; the
+     backend plan answers in its "Requests from the frontend lane" table —
+     published under an operation id, or declined with the reason. An answered
+     request is re-bound in the table above to the id the backend gave it; a
+     declined one is planned around, or routed to `specify` when the spec needs
+     it. Write `None` instead of the table when every binding is published. -->
+
+| Request | Operation id | Needed by | Sends (field shape) | Receives (field shape) | Serves |
+|---|---|---|---|---|---|
+| RQ-[n] | `[ProposedOperationId]` | `[Screen / component]` | [field: type, …] | [field: type, …] | FR-[nnn] |
 
 ## 6. Cross-Cutting
 

package/assets/.augment/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "augment",
+  "runtime": "Augment CLI (auggie --print)",
+  "note": "auggie has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: Augment rules and AGENTS.md still load. --print is the one-shot print flag; the prompt that follows it is the positional argument. Print mode skips the indexing confirmation and indexes the workspace root, which is where the engine runs. --quiet prints only the final assistant message. Read-only steps add --ask (retrieval and non-editing tools only); edit-capable steps run with Augment's own tool permissions. No sdk block: Augment has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "auggie",
+  "args": ["--quiet"],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "--print" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": [],
+    "bypassPermissions": [],
+    "default": [],
+    "readonly": ["--ask"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.augment/commands/appweaver.md

@@ -0,0 +1,12 @@
+---
+description: "Parametric launcher (Augment): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn)."
+argument-hint: "<skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API', 'analyze-slice'"
+---
+
+# AppWeaver launcher (Augment)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Augment: your row in its profile table is `.augment/agent.json`.
+
+The arguments are: $ARGUMENTS

package/assets/.cline/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "cline",
+  "runtime": "Cline CLI (cline)",
+  "note": "cline's -s/--system replaces Cline's own system prompt (its tool instructions), so 'system' is folded into the prompt instead (system: prepend). No tools allowlist or bare mode, so isolation is best-effort: .clinerules and AGENTS.md still load. The prompt is the positional argument, and -p means plan mode in Cline, so the engine puts '--' before the prompt. Steps run in act mode with --auto-approve true; read-only steps add --plan. Output stays plain text: --json prints NDJSON events, which the engine cannot read. -m takes a model id for the provider set up with 'cline auth'. The profile lives in .cline/ because Cline loads every file in .clinerules/ into every prompt as a rule; the launcher is .clinerules/workflows/appweaver.md. No sdk block: Cline has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "cline",
+  "args": ["--auto-approve", "true"],
+  "model_flag": "-m",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "--" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": [],
+    "bypassPermissions": [],
+    "default": [],
+    "readonly": ["--plan"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.clinerules/workflows/appweaver.md

@@ -0,0 +1,7 @@
+# AppWeaver launcher (Cline)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Cline: your row in its profile table is `.cline/agent.json`.
+
+The arguments are the text the user typed after `/appweaver.md`.

package/assets/.codex/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "codex",
+  "runtime": "OpenAI Codex CLI (codex exec)",
+  "note": "codex exec has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: AGENTS.md and Codex config still load. The prompt is read from stdin; progress goes to stderr and only the final agent message to stdout. Edit-capable steps run with --sandbox workspace-write: edits and commands inside the workspace, network off unless the developer's Codex config turns it on, so package restores can fail. Read-only steps run with --sandbox read-only. --skip-git-repo-check lets it run in a project that is not a git repository yet; --ephemeral keeps spawned steps out of the session history. -m takes a Codex model name. Codex has no launcher of its own: it reads .agents/skills/appweaver/SKILL.md. No sdk block: Codex has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "codex",
+  "args": ["exec", "--skip-git-repo-check", "--ephemeral", "--color", "never"],
+  "model_flag": "-m",
+  "system": "prepend",
+  "user": { "mode": "stdin" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--sandbox", "workspace-write"],
+    "bypassPermissions": ["--sandbox", "workspace-write"],
+    "default": [],
+    "readonly": ["--sandbox", "read-only"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.continue/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "continue",
+  "runtime": "Continue CLI (cn -p)",
+  "note": "cn has no system-prompt flag (--rule adds a rule, not a system prompt), so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: Continue rules still load. -p is a boolean print flag; the prompt that follows it is the positional argument. --silent strips think tags and excess whitespace from the reply. Edit-capable steps run with --auto (all tools allowed); read-only steps with --readonly (plan mode, read-only tools). --model takes a Continue Hub model slug (owner/package). No sdk block: Continue has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "cn",
+  "args": ["--silent"],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "-p" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--auto"],
+    "bypassPermissions": ["--auto"],
+    "default": [],
+    "readonly": ["--readonly"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.continue/prompts/appweaver.md

@@ -0,0 +1,13 @@
+---
+name: appweaver
+description: "Parametric launcher (Continue): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn). Usage: /appweaver <skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API'"
+invokable: true
+---
+
+# AppWeaver launcher (Continue)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Continue: your row in its profile table is `.continue/agent.json`.
+
+The arguments are the text the user typed after `/appweaver`.

package/assets/.cursor/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "cursor",
+  "runtime": "Cursor CLI (cursor-agent -p)",
+  "note": "cursor-agent has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: project rules such as .cursor/rules and AGENTS.md still load. 'agent' is the same binary, but other CLIs (e.g. Grok) also install 'agent', so the profile names cursor-agent. -p is a boolean print flag; the prompt that follows it is the positional argument. No sdk block: Cursor has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "cursor-agent",
+  "args": ["--output-format", "json", "--trust"],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "-p" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--force"],
+    "bypassPermissions": ["--force"],
+    "default": [],
+    "readonly": ["--mode", "ask"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.cursor/commands/appweaver.md

@@ -0,0 +1,7 @@
+# AppWeaver launcher (Cursor)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Cursor: your row in its profile table is `.cursor/agent.json`.
+
+The arguments are the text the user typed after `/appweaver`.

package/assets/.devin/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "devin",
+  "runtime": "Devin CLI (devin -p) — Devin Desktop, formerly Windsurf",
+  "note": "devin has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: project rules and AGENTS.md still load. Print mode takes the prompt as the value of -p (it does not read stdin) and cannot show the workspace-trust prompt, hence --respect-workspace-trust false. Edit-capable steps run with --permission-mode dangerous so they can also run builds and tests. Devin has no read-only mode flag: read-only steps run in its default 'normal' mode, which asks before edits; that print mode then refuses them is not yet confirmed on a real run. Output is plain text, so the engine reads raw stdout. No sdk block: Devin has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "devin",
+  "args": ["--respect-workspace-trust", "false"],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "-p" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--permission-mode", "dangerous"],
+    "bypassPermissions": ["--permission-mode", "dangerous"],
+    "default": [],
+    "readonly": []
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.devin/workflows/appweaver.md

@@ -0,0 +1,11 @@
+---
+description: "Parametric launcher (Devin Desktop): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn). Usage: /appweaver <skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API'"
+---
+
+# AppWeaver launcher (Devin Desktop)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Devin: your row in its profile table is `.devin/agent.json`.
+
+The arguments are the text the user typed after `/appweaver`.

package/assets/.factory/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "factory",
+  "runtime": "Factory Droid CLI (droid exec)",
+  "note": "droid exec reads the prompt from stdin and, with -o json, prints one object whose 'result' holds the reply. 'system' is folded into the prompt (system: prepend). No tools allowlist (droid's tool IDs differ from the engine's tool names) or bare mode, so isolation is best-effort: AGENTS.md and project droids still load. Without --auto, droid exec is read-only, so read-only steps pass no flag; edit-capable steps run with --auto high so they can also run builds, tests and git operations, including push. -m takes a Factory model id. No sdk block: Factory has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "droid",
+  "args": ["exec", "-o", "json"],
+  "model_flag": "-m",
+  "system": "prepend",
+  "user": { "mode": "stdin" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--auto", "high"],
+    "bypassPermissions": ["--auto", "high"],
+    "default": [],
+    "readonly": []
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.factory/commands/appweaver.md

@@ -0,0 +1,12 @@
+---
+description: "Parametric launcher (Factory Droid): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn)."
+argument-hint: "<skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API', 'analyze-slice'"
+---
+
+# AppWeaver launcher (Factory Droid)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Factory Droid: your row in its profile table is `.factory/agent.json`.
+
+The arguments are: $ARGUMENTS

package/assets/.gemini/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "gemini",
+  "runtime": "Gemini CLI (gemini -p)",
+  "note": "gemini has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: GEMINI.md and project Gemini settings still load. -p takes the prompt as its value. -o json prints one object whose 'response' holds the reply. --skip-trust trusts the workspace for the run, since headless mode cannot show the folder-trust prompt. Edit-capable steps run with --approval-mode auto_edit: file edits are auto-approved, shell commands are not, so builds and tests do not run inside spawned steps. Read-only steps run with --approval-mode plan. -m takes a Gemini model name. Gemini CLI has no launcher of its own: it reads .agents/skills/appweaver/SKILL.md. No sdk block: Gemini CLI has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "gemini",
+  "args": ["-o", "json", "--skip-trust"],
+  "model_flag": "-m",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "-p" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--approval-mode", "auto_edit"],
+    "bypassPermissions": ["--approval-mode", "auto_edit"],
+    "default": [],
+    "readonly": ["--approval-mode", "plan"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["response"]
+}

package/assets/.github/prompts/appweaver.prompt.md

@@ -0,0 +1,13 @@
+---
+description: "Parametric launcher (GitHub Copilot): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn)."
+argument-hint: "<skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API', 'analyze-slice'"
+agent: agent
+---
+
+# AppWeaver launcher (GitHub Copilot)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are GitHub Copilot: your row in its profile table is `.copilot/agent.json`.
+
+The arguments are the text the user typed after `/appweaver`.

package/assets/.kilo/agent.json

@@ -0,0 +1,28 @@
+{
+  "name": "kilocode",
+  "sdk": {
+    "type": "claude-agent-sdk",
+    "package": "claude-agent-sdk",
+    "module": "claude_agent_sdk",
+    "adapter": "../.appweaver/orchestrator/adapters/claude_agent_sdk.py"
+  },
+  "runtime": "Kilo CLI (kilo run)",
+  "note": "kilo run has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and the prompt is read from stdin (--prompt-stdin). No tools allowlist or bare mode, so isolation is best-effort: AGENTS.md and project Kilo rules still load. --auto refuses anything not explicitly allowed, so each step runs as one of the kit's Kilo agents in .kilo/agents/: appweaver-edit (read, edit, bash) or appweaver-read (read only). Output stays plain text: --format json prints JSON-lines events, which the engine cannot read. --model takes Kilo's provider/model form. Kilo has no Python SDK, so the sdk block uses the Claude Agent SDK: sdk/sdk-apply only work for developers who also have Claude access.",
+  "bin": "kilo",
+  "args": ["run", "--auto", "--prompt-stdin"],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "stdin" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--agent", "appweaver-edit"],
+    "bypassPermissions": ["--agent", "appweaver-edit"],
+    "default": ["--agent", "appweaver-edit"],
+    "readonly": ["--agent", "appweaver-read"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.kilo/agents/appweaver-edit.md

@@ -0,0 +1,12 @@
+---
+description: Runs one AppWeaver workflow step that may edit files and run commands (builds, tests). Used by the AppWeaver engine when it spawns `kilo run`; not for interactive use.
+mode: primary
+permission:
+  read: allow
+  glob: allow
+  grep: allow
+  edit: allow
+  bash: allow
+---
+
+You run one step of an AppWeaver workflow. The prompt holds your complete instructions and context: follow it exactly, and end your reply in the format it asks for.

package/assets/.kilo/agents/appweaver-read.md

@@ -0,0 +1,12 @@
+---
+description: Runs one read-only AppWeaver workflow step (author, query, review). Used by the AppWeaver engine when it spawns `kilo run`; not for interactive use.
+mode: primary
+permission:
+  read: allow
+  glob: allow
+  grep: allow
+  edit: deny
+  bash: deny
+---
+
+You run one read-only step of an AppWeaver workflow. The prompt holds your complete instructions and context: follow it exactly, do not modify any file, and end your reply in the format it asks for.

package/assets/.kilo/commands/appweaver.md

@@ -0,0 +1,11 @@
+---
+description: "Parametric launcher (Kilo Code): run ANY converted aw-spec skill on the aw-orchestrator under one of the 4 strategies (agent [default], spawn, sdk, sdk-apply). Usage: /appweaver <skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API'"
+---
+
+# AppWeaver launcher (Kilo Code)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Kilo Code: your row in its profile table is `.kilo/agent.json`.
+
+The arguments are: $ARGUMENTS

package/assets/.kiro/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "kiro",
+  "runtime": "Kiro CLI (kiro-cli chat --no-interactive)",
+  "note": "kiro-cli has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: Kiro steering and AGENTS.md still load. Headless mode reads the prompt from stdin. It cannot ask for tool approval, so edit-capable steps run with --trust-all-tools and read-only steps with --trust-tools=read,grep. Headless sign-in uses the KIRO_API_KEY environment variable. Output is plain text, so the engine reads raw stdout. No sdk block: Kiro has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "kiro-cli",
+  "args": ["chat", "--no-interactive"],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "stdin" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--trust-all-tools"],
+    "bypassPermissions": ["--trust-all-tools"],
+    "default": [],
+    "readonly": ["--trust-tools=read,grep"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.kiro/skills/appweaver/SKILL.md

@@ -0,0 +1,12 @@
+---
+name: appweaver
+description: "Parametric launcher (Kiro): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn). Usage: /appweaver <skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API'"
+---
+
+# AppWeaver launcher (Kiro)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Kiro: your row in its profile table is `.kiro/agent.json`.
+
+The arguments are the text the user typed after `/appweaver`.

package/assets/.opencode/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "opencode",
+  "runtime": "opencode CLI (opencode run)",
+  "note": "opencode run has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: AGENTS.md and project opencode rules still load. The message is positional only (no stdin), so the engine puts '--' before the prompt. --auto approves every permission that is not explicitly denied, so each step runs as one of the kit's opencode agents in .opencode/agents/: appweaver-edit (read, edit, bash) or appweaver-read (read only; edit, bash and subagents denied). Output stays plain text: --format json prints raw JSON events, which the engine cannot read. --model takes opencode's provider/model form. No sdk block: opencode has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "opencode",
+  "args": ["run", "--auto"],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "--" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--agent", "appweaver-edit"],
+    "bypassPermissions": ["--agent", "appweaver-edit"],
+    "default": ["--agent", "appweaver-edit"],
+    "readonly": ["--agent", "appweaver-read"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.opencode/agents/appweaver-edit.md

@@ -0,0 +1,12 @@
+---
+description: Runs one AppWeaver workflow step that may edit files and run commands (builds, tests). Used by the AppWeaver engine when it spawns `opencode run`; not for interactive use.
+mode: primary
+permission:
+  read: allow
+  glob: allow
+  grep: allow
+  edit: allow
+  bash: allow
+---
+
+You run one step of an AppWeaver workflow. The prompt holds your complete instructions and context: follow it exactly, and end your reply in the format it asks for.

package/assets/.opencode/agents/appweaver-read.md

@@ -0,0 +1,13 @@
+---
+description: Runs one read-only AppWeaver workflow step (author, query, review). Used by the AppWeaver engine when it spawns `opencode run`; not for interactive use.
+mode: primary
+permission:
+  read: allow
+  glob: allow
+  grep: allow
+  edit: deny
+  bash: deny
+  task: deny
+---
+
+You run one read-only step of an AppWeaver workflow. The prompt holds your complete instructions and context: follow it exactly, do not modify any file, and end your reply in the format it asks for.

package/assets/.opencode/commands/appweaver.md

@@ -0,0 +1,11 @@
+---
+description: "Parametric launcher (opencode): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn). Usage: /appweaver <skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API'"
+---
+
+# AppWeaver launcher (opencode)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are opencode: your row in its profile table is `.opencode/agent.json`.
+
+The arguments are: $ARGUMENTS

package/assets/.roo/agent.json

@@ -0,0 +1,22 @@
+{
+  "name": "roo",
+  "runtime": "Roo Code CLI (roo --print)",
+  "note": "roo has no system-prompt flag, so 'system' is folded into the prompt (system: prepend), and no tools allowlist or bare mode, so isolation is best-effort: .roo rules and AGENTS.md still load. --print is a boolean print flag; the prompt that follows it is the positional argument. Roo auto-approves tool use unless --require-approval is passed; edit-capable steps run in --mode code, read-only steps in --mode ask. Roo's default provider is OpenRouter: set your own provider and key (--provider / --api-key, or Roo's environment variables) before using spawn. Output is plain text, so the engine reads raw stdout. No sdk block: Roo has no Python SDK adapter, so sdk/sdk-apply are unavailable.",
+  "bin": "roo",
+  "args": [],
+  "model_flag": "--model",
+  "system": "prepend",
+  "user": { "mode": "arg", "flag": "--print" },
+  "tools_flag": null,
+  "approval": {
+    "acceptEdits": ["--mode", "code"],
+    "bypassPermissions": ["--mode", "code"],
+    "default": ["--mode", "code"],
+    "readonly": ["--mode", "ask"]
+  },
+  "isolation": {
+    "bare": [],
+    "fresh": []
+  },
+  "result_keys": ["result"]
+}

package/assets/.roo/commands/appweaver.md

@@ -0,0 +1,12 @@
+---
+description: "Parametric launcher (Roo Code): run ANY converted aw-spec skill on the aw-orchestrator under one of 2 strategies (agent [default], spawn)."
+argument-hint: "<skill> [strategy] [input…] — e.g. 'select-stack', 'specify spawn Build an order API', 'analyze-slice'"
+---
+
+# AppWeaver launcher (Roo Code)
+
+Read `.appweaver/orchestrator/LAUNCHER.md` and follow it exactly.
+
+You are Roo Code: your row in its profile table is `.roo/agent.json`.
+
+The arguments are the text the user typed after `/appweaver`.

package/dist/commands/init.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"init.d.ts","sourceRoot":"","sources":["../../src/commands/init.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAOpC,eAAO,MAAM,WAAW,SA2DpB,CAAC"}
+{"version":3,"file":"init.d.ts","sourceRoot":"","sources":["../../src/commands/init.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAOpC,eAAO,MAAM,WAAW,SAwEpB,CAAC"}

package/dist/commands/init.js

@@ -27,7 +27,20 @@ export const initCommand = new Command('init')
                     choices: [
                         { name: 'Claude Code (.claude/skills)', value: 'claude' },
                         { name: 'Antigravity (.agents/skills)', value: 'antigravity' },
-                        { name: 'KiloCode (.kilocode/skills)', value: 'kilocode' },
+                        { name: 'KiloCode (.kilocode/skills)', value: 'kilo' },
+                        { name: 'Cursor (.cursor/commands)', value: 'cursor' },
+                        { name: 'Cline (.clinerules/workflows)', value: 'cline' },
+                        { name: 'GitHub Copilot (.github/prompts)', value: 'github' },
+                        { name: 'Devin CLI (.devin/workflows)', value: 'devin' },
+                        { name: 'Roo code (.roo/commands)', value: 'roo' },
+                        { name: 'Opencode (.opencode/commands)', value: 'opencode' },
+                        { name: 'Factory Droid (.factory/commands)', value: 'factory' },
+                        { name: 'Augment (.augment/commands)', value: 'augment' },
+                        { name: 'Continue (.continue/prompts)', value: 'continue' },
+                        { name: 'Kiro (.kiro/skills)', value: 'kiro' },
+                        { name: 'OpenAI Codex (.agents/skills)', value: 'codex' },
+                        { name: 'Gemini CLI (.agents/skills)', value: 'gemini' },
+                        { name: 'Amp (.agents/skills)', value: 'amp' }
                     ],
                     default: 'claude',
                 },