code-agnostic 0.3.15__tar.gz → 0.3.17__tar.gz

{code_agnostic-0.3.15 → code_agnostic-0.3.17}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: code-agnostic
-Version: 0.3.15
+Version: 0.3.17
 Summary: Centralized hub for LLM coding config: MCP, skills, rules, and agents.
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
@@ -56,9 +56,10 @@ AI coding tools each want config in a different place and format. When you use m
 ~/.cursor/                        Compiled & synced for Cursor
 ~/.codex/                         Compiled & synced for Codex (or CODEX_HOME)
 ~/.claude.json and ~/.claude/      Compiled & synced for Claude Code
+~/.copilot/ and repo .github/      Compiled & synced for GitHub Copilot
 ```
 
-Each resource is cross-compiled to the target editor's native format. Rules become `.mdc` files for Cursor, `AGENTS.md` sections for OpenCode/Codex, and `CLAUDE.local.md` memory for Claude Code.
+Each resource is cross-compiled to the target editor's native format. Rules become `.mdc` files for Cursor, `AGENTS.md` sections for OpenCode/Codex/Copilot, and `CLAUDE.local.md` memory for Claude Code.
 
 Legacy single-file rules, `skills/<name>/SKILL.md`, and markdown agents are still supported for migration, but bundle directories are the preferred source format for new config.
 
@@ -77,8 +78,9 @@ Today the implementation is still mixed: some assets are compiled and some are s
 
 Workspace sync may generate repo-local outputs, but those outputs are not
 source. Project-local skill folders generated inside a repo, such as
-`.agents/skills` or `.opencode/skills`, are also generated outputs; install
-skills into managed project source first, then run `plan` / `apply`.
+`.agents/skills`, `.opencode/skills`, or `.github/skills`, are also generated
+outputs; install skills into managed project source first, then run `plan` /
+`apply`.
 
 ## Install
 
@@ -114,6 +116,7 @@ code-agnostic import apply -a codex
 code-agnostic apps enable -a cursor
 code-agnostic apps enable -a opencode
 code-agnostic apps enable -a claude
+code-agnostic apps enable -a copilot
 
 # Preview and apply
 code-agnostic validate
@@ -123,19 +126,19 @@ code-agnostic apply
 
 ## Editor compatibility
 
-| Feature | OpenCode | Cursor | Codex | Claude Code |
-|---------|:--------:|:------:|:-----:|:-----------:|
-| MCP sync | yes | yes | yes | yes |
-| Rules sync (cross-compiled) | yes | yes | yes | yes |
-| Skills sync | yes | yes | yes | yes |
-| Agents sync | yes | yes | yes | yes |
-| Workspace root `AGENTS.md` link | yes | yes | yes | yes |
-| Native repo config include for workspace `AGENTS.md` | yes | -- | -- | -- |
-| Repo/subdir gets shared workspace instructions today | yes | -- | yes | yes |
-| Nested `AGENTS.md` discovery | -- | yes | yes | -- |
-| Workspace propagation | yes | yes | yes | yes |
-| Import from | yes | yes | yes | yes |
-| Interactive import (TUI) | yes | yes | yes | yes |
+| Feature | OpenCode | Cursor | Codex | Claude Code | GitHub Copilot |
+|---------|:--------:|:------:|:-----:|:-----------:|:--------------:|
+| MCP sync | yes | yes | yes | yes | yes |
+| Rules sync (cross-compiled) | yes | yes | yes | yes | yes |
+| Skills sync | yes | yes | yes | yes | yes |
+| Agents sync | yes | yes | yes | yes | yes |
+| Workspace root `AGENTS.md` link | yes | yes | yes | yes | yes |
+| Native repo config include for workspace `AGENTS.md` | yes | -- | -- | -- | -- |
+| Repo/subdir gets shared workspace instructions today | yes | -- | yes | yes | yes |
+| Nested `AGENTS.md` discovery | -- | yes | yes | -- | yes |
+| Workspace propagation | yes | yes | yes | yes | yes |
+| Import from | yes | yes | yes | yes | yes |
+| Interactive import (TUI) | yes | yes | yes | yes | yes |
 
 `yes` means the resource type is synced for that editor. Some metadata is still
 target-specific or lossy; run `code-agnostic explain-lossiness` to see fields
@@ -143,7 +146,7 @@ that are omitted or rejected for a selected target.
 
 Cursor workspace propagation writes repo-local MCP, skills, and agents when those resources exist in the workspace source config.
 
-OpenCode workspace configs write project-root `opencode.json` files that include the shared workspace `AGENTS.md` natively via `instructions`, so repos under the workspace get both repo-local and shared workspace instructions. Codex repos receive workspace instructions through a generated `AGENTS.override.md`, which is added to each repo's `.git/info/exclude`. Claude Code receives workspace instructions through generated `CLAUDE.local.md` files, never by editing committed `CLAUDE.md`.
+OpenCode workspace configs write project-root `opencode.json` files that include the shared workspace `AGENTS.md` natively via `instructions`, so repos under the workspace get both repo-local and shared workspace instructions. Codex repos receive workspace instructions through a generated `AGENTS.override.md`, which is added to each repo's `.git/info/exclude`. Claude Code receives workspace instructions through generated `CLAUDE.local.md` files, never by editing committed `CLAUDE.md`. Copilot repo-shared MCP, skills, and agents are written under `.github/`.
 
 Cursor documents `AGENTS.md` support in project roots and subdirectories. `code-agnostic` does not copy or link the shared workspace `AGENTS.md` into child repos; Cursor will load `AGENTS.md` files that already exist in the opened project. Codex documents nested `AGENTS.md` discovery, but not a native config include for an extra workspace file.
 
@@ -157,14 +160,13 @@ Plan-then-apply workflow. Preview every change before it touches disk.
 code-agnostic validate              # check canonical source files
 code-agnostic plan -a cursor        # dry-run for one editor
 code-agnostic plan                   # dry-run for all
-code-agnostic apply                  # apply changes for all enabled editors
 code-agnostic status                 # check drift and disabled app states
 code-agnostic explain-lossiness      # show fields omitted or rejected per editor
 ```
 
 Bare `plan` and `apply` target every enabled editor; bare `status` also shows
-disabled app states. Use `-a codex`, `-a cursor`, `-a opencode`, or `-a claude`
-when you want one editor at a time.
+disabled app states. Use `-a codex`, `-a cursor`, `-a opencode`, `-a claude`, or
+`-a copilot` when you want one editor at a time.
 
 If managed outputs need repair after an apply, restore the active synced revision:
 
@@ -264,14 +266,19 @@ Global skills live under `~/.config/code-agnostic/skills`. Workspace-local
 skills live under `~/.config/code-agnostic/workspaces/<name>/skills` and can be
 inspected with `code-agnostic skills list -w <name>`. Project-local skills live
 under `~/.config/code-agnostic/projects/<name>/skills` and are generated into
-the registered project directory by `plan` / `apply`. Codex generated skill
-outputs are written to `~/.agents/skills`, while Codex agents and config remain
-under `CODEX_HOME` when set, defaulting to `~/.codex`. Claude Code generated
-skills and agents are written under `~/.claude/skills` and `~/.claude/agents`,
-with workspace/project copies under repo-local `.claude/skills` and
-`.claude/agents`.
-
-If a target app discovers user-created repo-local skill folders such as `.agents/skills`, `.opencode/skills`, or `.claude/skills`, treat those as unmanaged app inputs unless they were generated from `code-agnostic` project source. Workspace and project sync write only the exact generated paths recorded in their `.sync-state.json` files.
+the registered project directory by `plan` / `apply`. The current first-class
+list/remove commands cover global and workspace scopes; project skill inventory
+and removal are not supported yet, and `code-agnostic projects list` only shows
+a coarse `[skills]` marker today. Codex generated skill outputs are written to
+`~/.agents/skills`, while Codex agents and config remain under `CODEX_HOME` when
+set, defaulting to `~/.codex`. Claude Code generated skills and agents are
+written under `~/.claude/skills` and `~/.claude/agents`, with workspace/project
+copies under repo-local `.claude/skills` and `.claude/agents`. GitHub Copilot
+generated skills and agents are written under `~/.copilot/skills` and
+`~/.copilot/agents`, with repo-shared copies under `.github/skills` and
+`.github/agents`.
+
+If a target app discovers user-created repo-local skill folders such as `.agents/skills`, `.opencode/skills`, `.claude/skills`, or `.github/skills`, treat those as unmanaged app inputs unless they were generated from `code-agnostic` project source. Workspace and project sync write only the exact generated paths recorded in their `.sync-state.json` files.
 
 Planned convenience command:
 
@@ -285,7 +292,7 @@ first implementation slice.
 
 ### Workspaces
 
-Register workspace directories. Workspace rules are compiled into a canonical `AGENTS.md` at the workspace root. Repos keep their own repo-specific `AGENTS.md`; Codex receives the workspace rules through generated, git-excluded `AGENTS.override.md` files, while OpenCode workspace configs write project-root `opencode.json` files that reference the shared workspace file through `instructions`. Claude receives generated `CLAUDE.local.md` files and project MCP entries in `~/.claude.json["projects"][absolute_repo_path]["mcpServers"]`. Workspace source config, skills, and agents are propagated into repo-local generated paths for OpenCode, Cursor, Codex, and Claude; user-created target app skill folders remain unmanaged unless they were generated from managed source.
+Register workspace directories. Workspace rules are compiled into a canonical `AGENTS.md` at the workspace root. Repos keep their own repo-specific `AGENTS.md`; Codex receives the workspace rules through generated, git-excluded `AGENTS.override.md` files, while OpenCode workspace configs write project-root `opencode.json` files that reference the shared workspace file through `instructions`. Claude receives generated `CLAUDE.local.md` files and project MCP entries in `~/.claude.json["projects"][absolute_repo_path]["mcpServers"]`. Copilot receives repo-shared `.github/mcp.json`, `.github/skills`, and `.github/agents`. Workspace source config, skills, and agents are propagated into repo-local generated paths for OpenCode, Cursor, Codex, Claude, and Copilot; user-created target app skill folders remain unmanaged unless they were generated from managed source.
 
 Cursor propagation intentionally stays to repo-local MCP, skills, and agents; it does not copy the shared workspace `AGENTS.md` into child repos.
 
@@ -313,6 +320,7 @@ Migrate existing config from any supported editor into the hub.
 code-agnostic import plan -a codex
 code-agnostic import apply -a codex
 code-agnostic import plan -a claude
+code-agnostic import plan -a copilot
 code-agnostic import apply -a cursor --include mcp --on-conflict overwrite
 code-agnostic import plan -a codex -i    # interactive TUI picker
 ```
@@ -355,6 +363,7 @@ The compiler migration is documented in:
 - [x] Interactive TUI for import selection
 - [x] Claude Code support
 - [x] Project-scoped skill installs and sync
+- [ ] Project-scoped skill inventory and remove commands
 - [ ] `rules add` / `skills add` / `agents add` commands (open `$EDITOR` with template)
 - [ ] Shell auto-complete
 - [ ] Full TUI mode (command palette + menus)

{code_agnostic-0.3.15 → code_agnostic-0.3.17}/README.md

@@ -31,9 +31,10 @@ AI coding tools each want config in a different place and format. When you use m
 ~/.cursor/                        Compiled & synced for Cursor
 ~/.codex/                         Compiled & synced for Codex (or CODEX_HOME)
 ~/.claude.json and ~/.claude/      Compiled & synced for Claude Code
+~/.copilot/ and repo .github/      Compiled & synced for GitHub Copilot
 ```
 
-Each resource is cross-compiled to the target editor's native format. Rules become `.mdc` files for Cursor, `AGENTS.md` sections for OpenCode/Codex, and `CLAUDE.local.md` memory for Claude Code.
+Each resource is cross-compiled to the target editor's native format. Rules become `.mdc` files for Cursor, `AGENTS.md` sections for OpenCode/Codex/Copilot, and `CLAUDE.local.md` memory for Claude Code.
 
 Legacy single-file rules, `skills/<name>/SKILL.md`, and markdown agents are still supported for migration, but bundle directories are the preferred source format for new config.
 
@@ -52,8 +53,9 @@ Today the implementation is still mixed: some assets are compiled and some are s
 
 Workspace sync may generate repo-local outputs, but those outputs are not
 source. Project-local skill folders generated inside a repo, such as
-`.agents/skills` or `.opencode/skills`, are also generated outputs; install
-skills into managed project source first, then run `plan` / `apply`.
+`.agents/skills`, `.opencode/skills`, or `.github/skills`, are also generated
+outputs; install skills into managed project source first, then run `plan` /
+`apply`.
 
 ## Install
 
@@ -89,6 +91,7 @@ code-agnostic import apply -a codex
 code-agnostic apps enable -a cursor
 code-agnostic apps enable -a opencode
 code-agnostic apps enable -a claude
+code-agnostic apps enable -a copilot
 
 # Preview and apply
 code-agnostic validate
@@ -98,19 +101,19 @@ code-agnostic apply
 
 ## Editor compatibility
 
-| Feature | OpenCode | Cursor | Codex | Claude Code |
-|---------|:--------:|:------:|:-----:|:-----------:|
-| MCP sync | yes | yes | yes | yes |
-| Rules sync (cross-compiled) | yes | yes | yes | yes |
-| Skills sync | yes | yes | yes | yes |
-| Agents sync | yes | yes | yes | yes |
-| Workspace root `AGENTS.md` link | yes | yes | yes | yes |
-| Native repo config include for workspace `AGENTS.md` | yes | -- | -- | -- |
-| Repo/subdir gets shared workspace instructions today | yes | -- | yes | yes |
-| Nested `AGENTS.md` discovery | -- | yes | yes | -- |
-| Workspace propagation | yes | yes | yes | yes |
-| Import from | yes | yes | yes | yes |
-| Interactive import (TUI) | yes | yes | yes | yes |
+| Feature | OpenCode | Cursor | Codex | Claude Code | GitHub Copilot |
+|---------|:--------:|:------:|:-----:|:-----------:|:--------------:|
+| MCP sync | yes | yes | yes | yes | yes |
+| Rules sync (cross-compiled) | yes | yes | yes | yes | yes |
+| Skills sync | yes | yes | yes | yes | yes |
+| Agents sync | yes | yes | yes | yes | yes |
+| Workspace root `AGENTS.md` link | yes | yes | yes | yes | yes |
+| Native repo config include for workspace `AGENTS.md` | yes | -- | -- | -- | -- |
+| Repo/subdir gets shared workspace instructions today | yes | -- | yes | yes | yes |
+| Nested `AGENTS.md` discovery | -- | yes | yes | -- | yes |
+| Workspace propagation | yes | yes | yes | yes | yes |
+| Import from | yes | yes | yes | yes | yes |
+| Interactive import (TUI) | yes | yes | yes | yes | yes |
 
 `yes` means the resource type is synced for that editor. Some metadata is still
 target-specific or lossy; run `code-agnostic explain-lossiness` to see fields
@@ -118,7 +121,7 @@ that are omitted or rejected for a selected target.
 
 Cursor workspace propagation writes repo-local MCP, skills, and agents when those resources exist in the workspace source config.
 
-OpenCode workspace configs write project-root `opencode.json` files that include the shared workspace `AGENTS.md` natively via `instructions`, so repos under the workspace get both repo-local and shared workspace instructions. Codex repos receive workspace instructions through a generated `AGENTS.override.md`, which is added to each repo's `.git/info/exclude`. Claude Code receives workspace instructions through generated `CLAUDE.local.md` files, never by editing committed `CLAUDE.md`.
+OpenCode workspace configs write project-root `opencode.json` files that include the shared workspace `AGENTS.md` natively via `instructions`, so repos under the workspace get both repo-local and shared workspace instructions. Codex repos receive workspace instructions through a generated `AGENTS.override.md`, which is added to each repo's `.git/info/exclude`. Claude Code receives workspace instructions through generated `CLAUDE.local.md` files, never by editing committed `CLAUDE.md`. Copilot repo-shared MCP, skills, and agents are written under `.github/`.
 
 Cursor documents `AGENTS.md` support in project roots and subdirectories. `code-agnostic` does not copy or link the shared workspace `AGENTS.md` into child repos; Cursor will load `AGENTS.md` files that already exist in the opened project. Codex documents nested `AGENTS.md` discovery, but not a native config include for an extra workspace file.
 
@@ -132,14 +135,13 @@ Plan-then-apply workflow. Preview every change before it touches disk.
 code-agnostic validate              # check canonical source files
 code-agnostic plan -a cursor        # dry-run for one editor
 code-agnostic plan                   # dry-run for all
-code-agnostic apply                  # apply changes for all enabled editors
 code-agnostic status                 # check drift and disabled app states
 code-agnostic explain-lossiness      # show fields omitted or rejected per editor
 ```
 
 Bare `plan` and `apply` target every enabled editor; bare `status` also shows
-disabled app states. Use `-a codex`, `-a cursor`, `-a opencode`, or `-a claude`
-when you want one editor at a time.
+disabled app states. Use `-a codex`, `-a cursor`, `-a opencode`, `-a claude`, or
+`-a copilot` when you want one editor at a time.
 
 If managed outputs need repair after an apply, restore the active synced revision:
 
@@ -239,14 +241,19 @@ Global skills live under `~/.config/code-agnostic/skills`. Workspace-local
 skills live under `~/.config/code-agnostic/workspaces/<name>/skills` and can be
 inspected with `code-agnostic skills list -w <name>`. Project-local skills live
 under `~/.config/code-agnostic/projects/<name>/skills` and are generated into
-the registered project directory by `plan` / `apply`. Codex generated skill
-outputs are written to `~/.agents/skills`, while Codex agents and config remain
-under `CODEX_HOME` when set, defaulting to `~/.codex`. Claude Code generated
-skills and agents are written under `~/.claude/skills` and `~/.claude/agents`,
-with workspace/project copies under repo-local `.claude/skills` and
-`.claude/agents`.
-
-If a target app discovers user-created repo-local skill folders such as `.agents/skills`, `.opencode/skills`, or `.claude/skills`, treat those as unmanaged app inputs unless they were generated from `code-agnostic` project source. Workspace and project sync write only the exact generated paths recorded in their `.sync-state.json` files.
+the registered project directory by `plan` / `apply`. The current first-class
+list/remove commands cover global and workspace scopes; project skill inventory
+and removal are not supported yet, and `code-agnostic projects list` only shows
+a coarse `[skills]` marker today. Codex generated skill outputs are written to
+`~/.agents/skills`, while Codex agents and config remain under `CODEX_HOME` when
+set, defaulting to `~/.codex`. Claude Code generated skills and agents are
+written under `~/.claude/skills` and `~/.claude/agents`, with workspace/project
+copies under repo-local `.claude/skills` and `.claude/agents`. GitHub Copilot
+generated skills and agents are written under `~/.copilot/skills` and
+`~/.copilot/agents`, with repo-shared copies under `.github/skills` and
+`.github/agents`.
+
+If a target app discovers user-created repo-local skill folders such as `.agents/skills`, `.opencode/skills`, `.claude/skills`, or `.github/skills`, treat those as unmanaged app inputs unless they were generated from `code-agnostic` project source. Workspace and project sync write only the exact generated paths recorded in their `.sync-state.json` files.
 
 Planned convenience command:
 
@@ -260,7 +267,7 @@ first implementation slice.
 
 ### Workspaces
 
-Register workspace directories. Workspace rules are compiled into a canonical `AGENTS.md` at the workspace root. Repos keep their own repo-specific `AGENTS.md`; Codex receives the workspace rules through generated, git-excluded `AGENTS.override.md` files, while OpenCode workspace configs write project-root `opencode.json` files that reference the shared workspace file through `instructions`. Claude receives generated `CLAUDE.local.md` files and project MCP entries in `~/.claude.json["projects"][absolute_repo_path]["mcpServers"]`. Workspace source config, skills, and agents are propagated into repo-local generated paths for OpenCode, Cursor, Codex, and Claude; user-created target app skill folders remain unmanaged unless they were generated from managed source.
+Register workspace directories. Workspace rules are compiled into a canonical `AGENTS.md` at the workspace root. Repos keep their own repo-specific `AGENTS.md`; Codex receives the workspace rules through generated, git-excluded `AGENTS.override.md` files, while OpenCode workspace configs write project-root `opencode.json` files that reference the shared workspace file through `instructions`. Claude receives generated `CLAUDE.local.md` files and project MCP entries in `~/.claude.json["projects"][absolute_repo_path]["mcpServers"]`. Copilot receives repo-shared `.github/mcp.json`, `.github/skills`, and `.github/agents`. Workspace source config, skills, and agents are propagated into repo-local generated paths for OpenCode, Cursor, Codex, Claude, and Copilot; user-created target app skill folders remain unmanaged unless they were generated from managed source.
 
 Cursor propagation intentionally stays to repo-local MCP, skills, and agents; it does not copy the shared workspace `AGENTS.md` into child repos.
 
@@ -288,6 +295,7 @@ Migrate existing config from any supported editor into the hub.
 code-agnostic import plan -a codex
 code-agnostic import apply -a codex
 code-agnostic import plan -a claude
+code-agnostic import plan -a copilot
 code-agnostic import apply -a cursor --include mcp --on-conflict overwrite
 code-agnostic import plan -a codex -i    # interactive TUI picker
 ```
@@ -330,6 +338,7 @@ The compiler migration is documented in:
 - [x] Interactive TUI for import selection
 - [x] Claude Code support
 - [x] Project-scoped skill installs and sync
+- [ ] Project-scoped skill inventory and remove commands
 - [ ] `rules add` / `skills add` / `agents add` commands (open `$EDITOR` with template)
 - [ ] Shell auto-complete
 - [ ] Full TUI mode (command palette + menus)

{code_agnostic-0.3.15 → code_agnostic-0.3.17}/code_agnostic/__init__.py

@@ -1,3 +1,3 @@
 __all__ = ["__version__"]
 
-__version__ = "0.3.15"
+__version__ = "0.3.17"

code_agnostic-0.3.17/code_agnostic/agents/compilers.py

@@ -0,0 +1,176 @@
+"""Per-editor agent compilers."""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from typing import Any
+
+import yaml
+
+from code_agnostic.agents.claude import serialize_claude_agent
+from code_agnostic.agents.codex import serialize_codex_agent
+from code_agnostic.agents.models import Agent
+from code_agnostic.agents.opencode import serialize_opencode_agent
+from code_agnostic.errors import InvalidConfigSchemaError
+
+_CURSOR_AGENT_FRONTMATTER_KEYS = frozenset(
+    {"name", "description", "model", "readonly", "is_background"}
+)
+_COPILOT_AGENT_FRONTMATTER_KEYS = frozenset(
+    {
+        "name",
+        "description",
+        "model",
+        "target",
+        "disable-model-invocation",
+        "user-invocable",
+        "infer",
+        "mcp-servers",
+        "metadata",
+    }
+)
+
+
+class IAgentCompiler(ABC):
+    @abstractmethod
+    def compile(self, agent: Agent) -> str:
+        """Return compiled agent content for target editor."""
+
+
+class OpenCodeAgentCompiler(IAgentCompiler):
+    """Cross-compile for OpenCode agents."""
+
+    def compile(self, agent: Agent) -> str:
+        return serialize_opencode_agent(agent)
+
+
+class CursorAgentCompiler(IAgentCompiler):
+    """Cross-compile for Cursor subagents."""
+
+    def compile(self, agent: Agent) -> str:
+        return _serialize_cursor_agent(agent)
+
+
+def _serialize_cursor_agent(agent: Agent) -> str:
+    overrides = dict(agent.metadata.app_overrides.get("cursor", {}))
+    fm: dict[str, Any] = {}
+    if agent.metadata.name:
+        fm["name"] = agent.metadata.name
+    if agent.metadata.description:
+        fm["description"] = agent.metadata.description
+
+    model = overrides.pop("model", agent.metadata.model)
+    if model:
+        fm["model"] = model
+
+    if agent.metadata.tools.write is False:
+        fm["readonly"] = True
+
+    for key in sorted(overrides):
+        if key not in _CURSOR_AGENT_FRONTMATTER_KEYS:
+            allowed = ", ".join(sorted(_CURSOR_AGENT_FRONTMATTER_KEYS))
+            raise InvalidConfigSchemaError(
+                agent.source_path,
+                f"x-cursor.{key} is not supported in cursor agent "
+                f"frontmatter; allowed keys: {allowed}",
+            )
+        fm[key] = overrides[key]
+
+    parts: list[str] = []
+    if fm:
+        parts.append("---")
+        parts.append(yaml.dump(fm, default_flow_style=False, sort_keys=False).rstrip())
+        parts.append("---")
+        parts.append("")
+    parts.append(agent.content)
+    return "\n".join(parts)
+
+
+class CodexAgentCompiler(IAgentCompiler):
+    """Cross-compile for Codex subagents."""
+
+    def compile(self, agent: Agent) -> str:
+        return serialize_codex_agent(agent)
+
+
+class CopilotAgentCompiler(IAgentCompiler):
+    """Cross-compile for GitHub Copilot custom agents."""
+
+    def compile(self, agent: Agent) -> str:
+        return _serialize_copilot_agent(agent)
+
+
+def _serialize_copilot_agent(agent: Agent) -> str:
+    overrides = dict(agent.metadata.app_overrides.get("copilot", {}))
+    fm: dict[str, Any] = {}
+
+    name = overrides.pop("name", agent.metadata.name or agent.name)
+    if name:
+        fm["name"] = name
+
+    description = overrides.pop("description", agent.metadata.description)
+    if not description:
+        raise InvalidConfigSchemaError(
+            agent.source_path,
+            "Copilot agents require a description",
+        )
+    fm["description"] = description
+
+    model = overrides.pop("model", agent.metadata.model)
+    if model:
+        fm["model"] = model
+
+    tools = _copilot_tools(agent)
+    if tools is not None:
+        fm["tools"] = tools
+
+    for key in sorted(overrides):
+        if key not in _COPILOT_AGENT_FRONTMATTER_KEYS:
+            allowed = ", ".join(sorted(_COPILOT_AGENT_FRONTMATTER_KEYS))
+            raise InvalidConfigSchemaError(
+                agent.source_path,
+                f"x-copilot.{key} is not supported in copilot agent "
+                f"frontmatter; allowed keys: {allowed}",
+            )
+        if key in fm:
+            continue
+        fm[key] = overrides[key]
+
+    parts: list[str] = []
+    if fm:
+        parts.append("---")
+        parts.append(yaml.dump(fm, default_flow_style=False, sort_keys=False).rstrip())
+        parts.append("---")
+        parts.append("")
+    parts.append(agent.content)
+    return "\n".join(parts)
+
+
+def _copilot_tools(agent: Agent) -> list[str] | None:
+    permissions = agent.metadata.tools
+    if permissions.read is True and permissions.write is True and not permissions.mcp:
+        return None
+
+    tools: list[str] = []
+    if permissions.read:
+        tools.append("read")
+    if permissions.write:
+        tools.append("edit")
+
+    for item in agent.metadata.tools.mcp:
+        server = item.get("server")
+        tool = item.get("tool")
+        if not server:
+            raise InvalidConfigSchemaError(
+                agent.source_path,
+                "Copilot agent MCP tools require a server",
+            )
+        tools.append(f"{server}/{tool}" if tool else f"{server}/*")
+    return tools
+
+
+class ClaudeAgentCompiler(IAgentCompiler):
+    """Cross-compile for Claude Code subagents."""
+
+    def compile(self, agent: Agent) -> str:
+        return serialize_claude_agent(agent)

code_agnostic-0.3.17/code_agnostic/agents/copilot.py

@@ -0,0 +1,117 @@
+"""GitHub Copilot custom agent Markdown conversion."""
+
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+from code_agnostic.agents.models import Agent, AgentMetadata, AgentToolPermissions
+
+_FRONTMATTER_RE = re.compile(r"^---\s*\n(.*?)\n---\s*\n", re.DOTALL)
+_READ_ALIASES = {"read", "notebookread", "view"}
+_WRITE_ALIASES = {"edit", "multiedit", "write", "notebookedit"}
+
+
+@dataclass(frozen=True)
+class CopilotAgentParseResult:
+    agent: Agent
+    warnings: list[str]
+
+
+def parse_copilot_agent(path: Path) -> CopilotAgentParseResult:
+    text = path.read_text(encoding="utf-8")
+    raw: dict[str, Any]
+    match = _FRONTMATTER_RE.match(text)
+    if match:
+        raw = yaml.safe_load(match.group(1)) or {}
+        content = text[match.end() :]
+    else:
+        raw = {}
+        content = text
+    if not isinstance(raw, dict):
+        raw = {}
+
+    fallback_name = _copilot_agent_stem(path)
+    tools, warnings = _parse_copilot_tools(raw.get("tools"), path)
+    agent = Agent(
+        name=fallback_name,
+        source_path=path,
+        metadata=AgentMetadata(
+            name=str(raw.get("name", fallback_name)),
+            description=str(raw.get("description", "")),
+            model=str(raw.get("model", "")),
+            tools=tools,
+        ),
+        content=content,
+    )
+    return CopilotAgentParseResult(agent=agent, warnings=warnings)
+
+
+def copilot_agent_filename(path: Path) -> str:
+    return f"{_copilot_agent_stem(path)}.agent.md"
+
+
+def _copilot_agent_stem(path: Path) -> str:
+    name = path.name
+    if name.endswith(".agent.md"):
+        return name[: -len(".agent.md")]
+    if name.endswith(".md"):
+        return name[: -len(".md")]
+    return path.stem
+
+
+def _parse_copilot_tools(
+    raw: Any, path: Path
+) -> tuple[AgentToolPermissions, list[str]]:
+    if raw is None:
+        return AgentToolPermissions(), []
+
+    tokens = _coerce_tool_tokens(raw)
+    if tokens is None:
+        return AgentToolPermissions(), [
+            f"Skipped unrepresentable Copilot tools: {path}"
+        ]
+
+    if "*" in tokens:
+        return AgentToolPermissions(), []
+
+    read = False
+    write = False
+    mcp: list[dict[str, str]] = []
+    unsupported: list[str] = []
+    for token in tokens:
+        normalized = token.lower()
+        if normalized in _READ_ALIASES:
+            read = True
+            continue
+        if normalized in _WRITE_ALIASES:
+            write = True
+            continue
+        if "/" in token:
+            server, tool = token.split("/", 1)
+            if server and tool:
+                item = {"server": server}
+                if tool != "*":
+                    item["tool"] = tool
+                mcp.append(item)
+                continue
+        unsupported.append(token)
+
+    if unsupported:
+        return AgentToolPermissions(), [
+            f"Skipped unrepresentable Copilot tools in {path}: {', '.join(unsupported)}"
+        ]
+
+    return AgentToolPermissions(read=read, write=write, mcp=mcp), []
+
+
+def _coerce_tool_tokens(raw: Any) -> list[str] | None:
+    if isinstance(raw, str):
+        return [item.strip() for item in raw.split(",") if item.strip()]
+    if isinstance(raw, list):
+        return [item for item in raw if isinstance(item, str)]
+    return None

{code_agnostic-0.3.15 → code_agnostic-0.3.17}/code_agnostic/agents/parser.py

@@ -19,7 +19,7 @@ from code_agnostic.agents.models import (
 from code_agnostic.spec.loaders import load_agent_bundle
 
 _FRONTMATTER_RE = re.compile(r"^---\s*\n(.*?)\n---\s*\n", re.DOTALL)
-_APP_OVERRIDE_PREFIXES = ("cursor", "codex", "opencode", "claude")
+_APP_OVERRIDE_PREFIXES = ("cursor", "codex", "opencode", "claude", "copilot")
 
 
 def parse_agent(path: Path) -> Agent:

{code_agnostic-0.3.15 → code_agnostic-0.3.17}/code_agnostic/apps/app_id.py

@@ -6,6 +6,8 @@ from code_agnostic.constants import (
     CLAUDE_PROJECT_DIRNAME,
     CODEX_CONFIG_FILENAME,
     CODEX_PROJECT_DIRNAME,
+    COPILOT_PROJECT_CONFIG_FILENAME,
+    COPILOT_PROJECT_DIRNAME,
     CURSOR_CONFIG_FILENAME,
     CURSOR_PROJECT_DIRNAME,
     OPENCODE_CONFIG_FILENAME,
@@ -19,6 +21,7 @@ class AppId(str, Enum):
     CURSOR = "cursor"
     CODEX = "codex"
     CLAUDE = "claude"
+    COPILOT = "copilot"
 
 
 @dataclass(frozen=True)
@@ -89,6 +92,17 @@ APP_CATALOG: dict[AppId, AppMetadata] = {
         project_dir_name=CLAUDE_PROJECT_DIRNAME,
         config_filename=CLAUDE_CONFIG_FILENAME,
     ),
+    AppId.COPILOT: AppMetadata(
+        app_id=AppId.COPILOT,
+        label="GitHub Copilot",
+        targetable=True,
+        toggleable=True,
+        importable=True,
+        supports_import_agents=True,
+        supports_workspace_propagation=True,
+        project_dir_name=COPILOT_PROJECT_DIRNAME,
+        config_filename=COPILOT_PROJECT_CONFIG_FILENAME,
+    ),
 }