feat(agents): add GitHub Copilot agent provider

CHANGELOG.md

@@ -13,8 +13,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- **F089**: GitHub Copilot agent provider — new `github_copilot` provider implementing `AgentProvider` interface via the `copilot` CLI binary; single-turn execution (`-p <prompt> --output-format=json --silent`), multi-turn conversation support (`--resume=<session-id>`), CLI flag mapping for model (`--model`), mode (`--mode`: interactive/plan/autopilot), effort (`--effort`: low/medium/high), tool permissions (`--allow-tool`, `--deny-tool`, `--allow-all`); JSONL streaming display event parsing; option validation for `mode` and `effort` enums at `awf validate` time; graceful fallback to stateless mode when session ID extraction fails; registered in `RegisterDefaults()` as 6th provider; authenticate via `copilot login` or environment variables (`COPILOT_GITHUB_TOKEN`, `GH_TOKEN`)
 - **F088**: Terminal User Interface (TUI) — `awf tui` launches a full-screen Bubble Tea dashboard with five tabs: Workflows (filterable list with launch/validate actions), Monitoring (real-time execution tree with status icons and live log viewport via 200ms tick polling of `ExecutionContext`), History (browse past executions from SQLite with filtering by name/status/date), Agent Conversations (Glamour-rendered Markdown chat view), and External Logs (fsnotify-based live tailing of Claude Code JSONL session files); TUI lives in `internal/interfaces/tui/` as a new interface adapter bridging to existing `WorkflowService`, `ExecutionService`, and `HistoryService` via async `tea.Cmd` factories; secret masking applied to all views; terminal state restored on exit/panic/signal; requires 256-color terminal support (graceful fallback to basic ANSI)
-- **F085**: Unified display-event abstraction across all agent providers — replaces per-provider `LineExtractor` function-field with a `DisplayEventParser` returning structured `DisplayEvent` values (discriminated by `EventText` and `EventToolUse` kinds); all 5 providers (Claude, Codex, Gemini, OpenCode, OpenAI-Compatible) now emit events through the same parser contract; single interfaces-layer `RenderEvents` renderer with two display modes: default (text only, byte-equivalent to F082 behaviour) and verbose (text + tool-use markers in `[tool: Name(Arg)]` format); well-known tools (`Read`, `Write`, `Edit`, `Bash`, `Grep`, `Glob`, `Task`) display concise markers with argument truncation (≤ 40 chars); unknown tool names degrade gracefully; parser implementations return plain strings with no ANSI escapes (rendering concerns confined to interfaces layer); `output_format: json` bypasses event parsing entirely for raw passthrough; `DisplayOutput` aggregation on `AgentResult`/`ConversationResult` preserved via text-event concatenation
+- **F085**: Unified display-event abstraction across all agent providers — replaces per-provider `LineExtractor` function-field with a `DisplayEventParser` returning structured `DisplayEvent` values (discriminated by `EventText` and `EventToolUse` kinds); all 6 providers (Claude, Codex, Gemini, GitHub Copilot, OpenCode, OpenAI-Compatible) now emit events through the same parser contract; single interfaces-layer `RenderEvents` renderer with two display modes: default (text only, byte-equivalent to F082 behaviour) and verbose (text + tool-use markers in `[tool: Name(Arg)]` format); well-known tools (`Read`, `Write`, `Edit`, `Bash`, `Grep`, `Glob`, `Task`) display concise markers with argument truncation (≤ 40 chars); unknown tool names degrade gracefully; parser implementations return plain strings with no ANSI escapes (rendering concerns confined to interfaces layer); `output_format: json` bypasses event parsing entirely for raw passthrough; `DisplayOutput` aggregation on `AgentResult`/`ConversationResult` preserved via text-event concatenation
 
 ### Changed
 

CLAUDE.md

@@ -72,7 +72,7 @@ Use `mcp__plugin_common_serena__write_memory` or `mcp__plugin_common_serena__edi
 
 ## Project Overview
 
-**ai-workflow-cli** (`awf`) - A Go CLI tool for orchestrating AI agents (Claude, Gemini, Codex) through YAML-configured workflows with state machine execution.
+**ai-workflow-cli** (`awf`) - A Go CLI tool for orchestrating AI agents (Claude, Gemini, Codex, GitHub Copilot) through YAML-configured workflows with state machine execution.
 
 ## Architecture
 
@@ -217,7 +217,6 @@ func TestWorkflowValidation(t *testing.T) {
 
 ## Architecture Rules
 
-- Pass optional turn-specific configuration (e.g., system_prompt) through options map in application layer; keeps infrastructure providers independent of turn logic
 - Validate agent provider options only against what each CLI actually accepts; do not validate against API documentation if the underlying CLI rejects the option
 - Plugin binaries must be discoverable at <plugins_dir>/<plugin_name>/awf-plugin-<plugin_name>; host validates binary existence and version compatibility via gRPC handshake after process start
 - Commit generated protobuf files (.pb.go, _grpc.pb.go) to git; treat as source artifacts for build reproducibility, not ephemeral build outputs
@@ -240,6 +239,8 @@ func TestWorkflowValidation(t *testing.T) {
 - Wire optional render callbacks alongside event parsers in stream processors; decouples rendering from parsing and enables multiple render modes (DefaultMode, VerboseMode) without modifying parser implementations
 - When integrating external UI frameworks, create Bridge adapters in the interface layer that wrap application services; maintain zero infrastructure imports in bridge implementation
 
+- Use provider name prefixes for all infrastructure provider helper methods (buildCopilot, extractCopilot, parseCopilot, validateCopilot) to prevent naming collisions across implementations
+
 ## Common Pitfalls
 
 - Use 0o755 for executable scripts, 0o644 for data files, 0o700 for private temp files; match permissions to file purpose and access expectations

README.md

@@ -3,17 +3,17 @@
 [![Go Version](https://img.shields.io/badge/Go-1.24+-00ADD8?style=flat&logo=go)](https://go.dev/)
 [![License: EUPL-1.2](https://img.shields.io/badge/License-EUPL--1.2-blue.svg)](LICENSE)
 
-A Go CLI tool for orchestrating AI agents (Claude, Gemini, Codex, OpenAI-Compatible APIs) through YAML-configured workflows with state machine execution.
+A Go CLI tool for orchestrating AI agents (Claude, Gemini, Codex, GitHub Copilot, OpenAI-Compatible APIs) through YAML-configured workflows with state machine execution.
 
 ## Features
 
 - **State Machine Execution** - Define workflows as state machines with conditional transitions based on exit codes, command output, or custom expressions
 - **Inline Error Handling** - Specify error messages and exit codes directly on steps without creating separate terminal states
-- **Agent Steps** - Invoke AI agents via CLI tools (Claude, Codex, Gemini) or direct HTTP (OpenAI, Ollama, vLLM, Groq) with prompt templates, response parsing, and accurate token tracking
-- **Output Formatting for Agent Steps** - Automatically strip markdown code fences and validate JSON output; human-readable streaming display controlled by `output_format` field (text vs raw NDJSON); unified display-event abstraction across all 5 providers with optional verbose mode showing tool-use markers (`[tool: Name(Arg)]`)
+- **Agent Steps** - Invoke AI agents via CLI tools (Claude, Codex, Gemini, GitHub Copilot) or direct HTTP (OpenAI, Ollama, vLLM, Groq) with prompt templates, response parsing, and accurate token tracking
+- **Output Formatting for Agent Steps** - Automatically strip markdown code fences and validate JSON output; human-readable streaming display controlled by `output_format` field (text vs raw NDJSON); unified display-event abstraction across all 6 providers with optional verbose mode showing tool-use markers (`[tool: Name(Arg)]`)
 - **External Prompt Files** - Load agent prompts from `.md` files with full template interpolation, helper functions, and local override support
 - **External Script Files** - Load commands from external script files with shebang-based interpreter dispatch, template interpolation, path resolution, and local override support
-- **Conversation Mode** - Multi-turn conversations with native session resume for CLI providers (`claude`, `codex`, `gemini`, `opencode`), automatic context window management for HTTP providers, mid-conversation context injection via `inject_context` field, and token tracking across all turns
+- **Conversation Mode** - Multi-turn conversations with native session resume for CLI providers (`claude`, `codex`, `gemini`, `opencode`, `github_copilot`), automatic context window management for HTTP providers, mid-conversation context injection via `inject_context` field, and token tracking across all turns
 - **OpenAI-Compatible Provider** - Use any Chat Completions API (OpenAI, Ollama, vLLM, Groq) with native HTTP integration, accurate token reporting, and no CLI tool required
 - **Parallel Execution** - Run multiple steps concurrently with configurable strategies
 - **Loop Constructs** - For-each and while loops with full context access

docs/README.md

@@ -28,7 +28,7 @@ Learn how to use AWF effectively:
 
 - [Commands](user-guide/commands.md) - All CLI commands and flags
 - [Interactive Input Collection](user-guide/interactive-inputs.md) - Automatic prompting for missing workflow inputs
-- [Agent Steps](user-guide/agent-steps.md) - Invoke AI agents via CLI (Claude, Codex, Gemini) or HTTP APIs (OpenAI, Ollama, vLLM, Groq)
+- [Agent Steps](user-guide/agent-steps.md) - Invoke AI agents via CLI (Claude, Codex, Gemini, GitHub Copilot) or HTTP APIs (OpenAI, Ollama, vLLM, Groq)
   - [Output Formatting](user-guide/agent-steps.md#output-formatting) - Automatic code fence stripping and JSON validation (`output_format: json|text`)
   - [Streaming Output Display & Tool Markers](user-guide/agent-steps.md#streaming-output-display--tool-markers) - Human-readable filtered output and tool-use markers for `--output streaming` and `--output buffered` modes
   - [External Prompt Files](user-guide/agent-steps.md#external-prompt-files) - Load prompts from Markdown files with template interpolation

docs/user-guide/agent-steps.md

@@ -2,7 +2,7 @@
 title: "Agent Steps Guide"
 ---
 
-Invoke AI agents (Claude, Codex, Gemini, OpenCode, OpenAI-Compatible) in your workflows with structured prompts and response parsing.
+Invoke AI agents (Claude, Codex, Gemini, GitHub Copilot, OpenCode, OpenAI-Compatible) in your workflows with structured prompts and response parsing.
 
 ## Overview
 
@@ -126,6 +126,38 @@ summarize:
 - `model`: Gemini model identifier — validated to start with `gemini-` prefix (see [Model Validation](#model-validation) below)
 - `dangerously_skip_permissions`: Skip permission prompts (boolean, maps to `--approval-mode=yolo`). **Security warning**: bypasses all safety prompts — use only in trusted, automated environments.
 
+### GitHub Copilot
+
+Requires the `copilot` CLI tool installed and authentication via `copilot login` or environment variables.
+
+```yaml
+code_generate:
+  type: agent
+  provider: github_copilot
+  prompt: "Generate a function to: {{.inputs.requirement}}"
+  options:
+    model: gpt-4o
+    mode: interactive
+  timeout: 60
+  on_success: next
+```
+
+**Provider-Specific Options:**
+- `model`: GitHub Copilot model identifier (e.g., `gpt-4o`, `gpt-4`, `gpt-3.5-turbo`)
+- `mode`: Agent mode — one of `interactive` (default), `plan`, or `autopilot`
+- `effort`: Reasoning effort level — one of `low`, `medium`, or `high`
+- `allowed_tools`: Comma-separated list of tools to allow (e.g., `"bash,github_api"` → `--allow-tool bash --allow-tool github_api`)
+- `denied_tools`: Comma-separated list of tools to deny (maps to `--deny-tool`)
+- `allow_all`: Allow all available tools (boolean, maps to `--allow-all`)
+- `system_prompt`: Custom system message (passed via prompt prepending)
+
+**Authentication:**
+GitHub Copilot CLI supports authentication via:
+- `copilot login` (interactive authentication)
+- `COPILOT_GITHUB_TOKEN` environment variable
+- `GH_TOKEN` environment variable
+- `GITHUB_TOKEN` environment variable (classic PATs not supported)
+
 ### OpenCode
 
 Requires the `opencode` CLI tool installed.
@@ -242,9 +274,9 @@ options:
 step validation error: model must start with "gpt-", "codex-", or match o-series pattern (e.g., o1, o3-mini)
 ```
 
-### OpenCode & OpenAI-Compatible
+### GitHub Copilot, OpenCode & OpenAI-Compatible
 
-No model validation for `opencode` or `openai_compatible` providers — these use arbitrary backend models.
+No model validation for `github_copilot`, `opencode`, or `openai_compatible` providers — these support arbitrary backend models.
 
 ### When Validation Occurs
 
@@ -699,7 +731,7 @@ Tool markers show:
 - **Interleaved order** — markers appear in the same source order as agent output
 - **Graceful degradation** — unknown tool names display as-is with no crash or error
 
-This works consistently across all 5 supported providers (Claude, Codex, Gemini, OpenCode, OpenAI-Compatible). Verbose mode has no effect on `output_format: json` — raw NDJSON is always passed through unchanged.
+This works consistently across all 6 supported providers (Claude, Codex, Gemini, GitHub Copilot, OpenCode, OpenAI-Compatible). Verbose mode has no effect on `output_format: json` — raw NDJSON is always passed through unchanged.
 
 #### Buffered Mode (`--output buffered`)
 

docs/user-guide/conversation-steps.md

@@ -66,7 +66,7 @@ You'll see the agent's first reply, then a `> ` prompt where you can type. When
 
 | Field | Required | Description |
 |---|---|---|
-| `provider` | Yes | Agent provider (`claude`, `gemini`, `codex`, `opencode`, `openai_compatible`) |
+| `provider` | Yes | Agent provider (`claude`, `gemini`, `codex`, `opencode`, `github_copilot`, `openai_compatible`) |
 | `mode` | Yes | Must be `conversation` |
 | `prompt` | Yes | First user message — sent automatically as turn 1 |
 | `system_prompt` | No | System message preserved for the whole session |
@@ -137,7 +137,7 @@ states:
 Both steps run as `mode: single` (the default — no `mode:` line needed). There is no interactive loop. Each step runs exactly one agent turn.
 
 - `seed` has `conversation: {}`. This marks the step as session-tracked: AWF calls `provider.ExecuteConversation` (instead of `provider.Execute`), the provider runs one turn, and the session ID returned by the CLI is captured into `state.conversation.session_id`.
-- `recall` has `conversation: {continue_from: seed}`. AWF clones the conversation state from `seed` (session ID + turn history) and passes it to the provider, which resumes the session via its native flag (`claude -r <id>`, `gemini --resume <id>`, `codex resume <id>`, `opencode -s <id>`).
+- `recall` has `conversation: {continue_from: seed}`. AWF clones the conversation state from `seed` (session ID + turn history) and passes it to the provider, which resumes the session via its native flag (`claude -r <id>`, `gemini --resume <id>`, `codex resume <id>`, `opencode -s <id>`, `copilot --resume=<id>`).
 
 ### Why the Empty `conversation: {}`?