agent attribution v3

AGENTS.md

@@ -26,7 +26,7 @@ Agents are the primary entity. Each agent bundles 5 component types: MCP servers
 - Pi
 
 **Functional** (config gen and scanning work, but no session parser or hook spec):
-- Gemini CLI, Codex CLI, Copilot, Copilot CLI, OpenCode
+- Codex CLI, Copilot, Copilot CLI, OpenCode
 
 See `docs/adding-a-harness.md` for the complete guide to adding or promoting a harness.
 

docs/cli/pull.md

@@ -50,7 +50,7 @@ For Kiro:
 | `.kiro/steering/<name>.md` | Steering file (the agent's system instructions) |
 | `.kiro/settings/mcp.json` | MCP servers, wrapped via `observal-shim` |
 
-For Cursor / VS Code / Gemini CLI: primarily MCP config + rules files at the appropriate path.
+For Cursor, VS Code, Copilot, and OpenCode: primarily MCP config plus rules or agent files at the appropriate path.
 
 ## Environment variables
 

docs/cli/skill.md

@@ -136,7 +136,6 @@ Two scopes are supported:
 observal skill install my-skill --harness claude-code
 observal skill install @sk --harness kiro --scope project
 observal skill install 2 --harness cursor --raw > config.json
-observal skill install my-skill --harness gemini-cli --no-write
 ```
 
 | Option | Description |

docs/e2e-test-checklist.md

@@ -127,7 +127,6 @@
   - [ ] Kiro
   - [ ] Claude Code
   - [ ] Codex
-  - [ ] Gemini CLI
   - [ ] Copilot
   - [ ] Open Code
 

docs/getting-started/quickstart.md

@@ -92,7 +92,7 @@ observal auth whoami
 
 ## 4. Discover and instrument your harness
 
-If you already have MCP servers configured in Claude Code, Kiro, Cursor, VS Code, or Gemini CLI, first see what's there:
+If you already have MCP servers configured in Claude Code, Kiro, Cursor, VS Code, or Copilot, first see what's there:
 
 ```bash
 observal scan

docs/hooks.md

@@ -83,9 +83,6 @@ observal registry hook install <name> --harness cursor
 # Install for Kiro
 observal registry hook install <name> --harness kiro
 
-# Install for Gemini CLI
-observal registry hook install <name> --harness gemini-cli
-
 # Install into a specific directory
 observal registry hook install <name> --harness claude-code --dir /path/to/project
 
@@ -209,13 +206,13 @@ observal agent pull safe-coder --harness claude-code
 
 The registry maps canonical event names to each harness's format:
 
-| Observal Event | Claude Code | Cursor | Kiro | Gemini CLI | Codex CLI |
-|----------------|-------------|--------|------|-----------|-----------|
-| `PreToolUse` | `PreToolUse` | `preToolUse` | `preToolUse` | `BeforeTool` | `pre_tool_use` |
-| `PostToolUse` | `PostToolUse` | `postToolUse` | `postToolUse` | `AfterTool` | `post_tool_use` |
-| `Stop` | `Stop` | `sessionEnd` | `stop` | `SessionEnd` | `session_stop` |
-| `SessionStart` | `SessionStart` | `sessionStart` | `agentSpawn` | `SessionStart` | - |
-| `UserPromptSubmit` | `UserPromptSubmit` | `beforeSubmitPrompt` | `userPromptSubmit` | `BeforeAgent` | `user_prompt_submit` |
+| Observal Event | Claude Code | Cursor | Kiro | Codex CLI |
+|----------------|-------------|--------|------|-----------|
+| `PreToolUse` | `PreToolUse` | `preToolUse` | `preToolUse` | `pre_tool_use` |
+| `PostToolUse` | `PostToolUse` | `postToolUse` | `postToolUse` | `post_tool_use` |
+| `Stop` | `Stop` | `sessionEnd` | `stop` | `session_stop` |
+| `SessionStart` | `SessionStart` | `sessionStart` | `agentSpawn` | - |
+| `UserPromptSubmit` | `UserPromptSubmit` | `beforeSubmitPrompt` | `userPromptSubmit` | `user_prompt_submit` |
 
 Install generates the correct format automatically:
 
@@ -224,7 +221,6 @@ Install generates the correct format automatically:
 observal registry hook install my-hook --harness claude-code  # → .claude/settings.json
 observal registry hook install my-hook --harness cursor       # → .cursor/hooks.json
 observal registry hook install my-hook --harness kiro         # → ~/.kiro/agents/my-hook.json
-observal registry hook install my-hook --harness gemini-cli   # → .gemini/settings.json
 ```
 
 ## Timeout Enforcement

docs/reference/config-files.md

@@ -85,13 +85,6 @@ Use anywhere that accepts `<id-or-name>` by prefixing with `@`.
 | `.vscode/mcp.json` | MCP servers (wrapped via `observal-shim`) |
 | `AGENTS.md` | Rules loaded into context |
 
-### Gemini CLI
-
-| Path | Purpose |
-| --- | --- |
-| `.gemini/settings.json` | MCP servers + config |
-| `AGENTS.md` / `GEMINI.md` | Rules |
-
 ### Codex CLI
 
 | Path | Purpose |

docs/self-hosting/troubleshooting.md

@@ -138,7 +138,6 @@ cat ~/.kiro/settings/mcp.json       # Kiro global
 cat ~/.claude/settings.json         # Claude Code
 cat .cursor/mcp.json                # Cursor
 cat .vscode/mcp.json                # VS Code
-cat .gemini/settings.json           # Gemini CLI
 ```
 
 If none exist, configure at least one MCP server in your harness first, then re-run `doctor patch`.

docs/use-cases/observe-mcp-traffic.md

@@ -21,7 +21,7 @@ Spans stream into ClickHouse in near real-time. You query them from the web UI o
 
 ## Discover and instrument an existing setup
 
-If you already have MCP servers configured in Claude Code, Kiro, Cursor, VS Code, or Gemini CLI, first see what's there:
+If you already have MCP servers configured in Claude Code, Kiro, Cursor, VS Code, or Copilot, first see what's there:
 
 ```bash
 observal scan
@@ -35,7 +35,7 @@ observal doctor patch --all --all-harnesses
 
 This:
 
-1. Finds every MCP config file on your machine (`~/.claude/settings.json`, `.kiro/settings/mcp.json`, `.cursor/mcp.json`, `.vscode/mcp.json`, `.gemini/settings.json`).
+1. Finds every MCP config file on your machine (`~/.claude/settings.json`, `.kiro/settings/mcp.json`, `.cursor/mcp.json`, `.vscode/mcp.json`, `~/.copilot/mcp-config.json`).
 2. Rewrites each config so every server runs through `observal-shim`.
 3. Installs telemetry hooks for session lifecycle events.
 4. Saves a timestamped `.bak` next to every file it modified.
@@ -45,7 +45,7 @@ Scope to specific harnesses:
 ```bash
 observal doctor patch --all --harness claude-code
 observal doctor patch --all --harness kiro
-observal doctor patch --all --harness gemini-cli
+observal doctor patch --all --harness copilot-cli
 ```
 
 ## Observability at zero cost to your agents

docs/use-cases/share-agent-configs.md

@@ -60,7 +60,6 @@ Install with one command, pick the harness:
 observal agent pull <agent-id> --harness claude-code
 observal agent pull <agent-id> --harness kiro
 observal agent pull <agent-id> --harness cursor
-observal agent pull <agent-id> --harness gemini-cli
 observal agent pull <agent-id> --harness vscode
 observal agent pull <agent-id> --harness codex
 ```

observal-server/services/agent_builder.py

@@ -11,7 +11,6 @@
 Generates harness-specific agent files from a ResolvedAgent:
 - Claude Code: .claude/agents/<name>.md (markdown) + MCP JSON config
 - Cursor: .cursor/agents/<name>.md (subagent markdown) + .cursor/mcp.json
-- Gemini CLI: GEMINI.md (markdown) + MCP JSON config
 - Kiro: ~/.kiro/agents/<name>.json (JSON)
 - Codex: ~/.codex/agents/<name>.toml (custom agent)
 - GitHub Copilot: .github/copilot-instructions.md (markdown)

observal_cli/README.md

@@ -103,7 +103,7 @@ observal uninstall                                   # tear down Docker, remove
 | --------------------------- | ---------------------------- |
 | Claude Code                 | Fully supported              |
 | Kiro CLI                    | Supported (next most tested) |
-| Cursor, VS Code, Gemini CLI | Untested                     |
+| Cursor, VS Code             | Untested                     |
 
 The `--harness` flag controls which config format is generated. Each harness has its own config paths and JSON structure.
 

observal_cli/skills/observal-advanced/SKILL.md

@@ -62,7 +62,6 @@ Use **only** when a command exits with `Connection failed` or `Not configured`.
 | Claude Code | `~/.claude/agents/<name>.md` | `.claude/agents/<name>.md` |
 | Kiro | `~/.kiro/agents/<name>.json` | `.kiro/agents/<name>.json` |
 | Cursor | `~/.cursor/rules/<name>.mdc` | `.cursor/rules/<name>.mdc` |
-| Gemini CLI | `~/.gemini/agents/<name>.md` | `.gemini/agents/<name>.md` |
 | VS Code | `~/.config/Code/User/agents/<name>.md` | `.vscode/agents/<name>.md` |
 | Codex CLI | `~/.codex/agents/<name>.md` | `.codex/agents/<name>.md` |
 | Copilot CLI | `~/.config/github-copilot/agents/<name>.md` | `.github/copilot/agents/<name>.md` |
@@ -74,7 +73,7 @@ Use **only** when a command exits with `Connection failed` or `Not configured`.
 {"name":"<name>","description":"<desc>","prompt":"<prompt>","model":"claude-sonnet-4-20250514","mcpServers":{},"tools":["*"],"resources":["skill://~/.kiro/skills/*/SKILL.md"]}
 ```
 
-**Claude Code, Gemini CLI, VS Code, Codex CLI, Copilot CLI, OpenCode** (markdown):
+**Claude Code, VS Code, Codex CLI, Copilot CLI, OpenCode** (markdown):
 
 ```markdown
 ---

observal_cli/skills/observal/SKILL.md

@@ -35,9 +35,9 @@ observal agent pull AGENT_NAME --harness kiro --no-prompt --dir .
 ```
 
 **Flags:**
-- `--harness` (required): `claude-code`, `kiro`, `cursor`, `gemini-cli`, `vscode`, `codex`, `copilot`, `copilot-cli`, `opencode`, `antigravity`, `pi`
+- `--harness` (required): `claude-code`, `kiro`, `cursor`, `vscode`, `codex`, `copilot`, `copilot-cli`, `opencode`, `antigravity`, `pi`
 - `--version <semver>`: install a specific version (e.g. `1.2.0`). Omit for latest.
-- `--scope user|project`: install scope (Claude Code, Kiro, Gemini only)
+- `--scope user|project`: install scope for harnesses that support user or project installs
 - `--model <name>` or `--model <harness>=<name>`: override saved model (repeatable)
 - `--tools t1,t2`: Claude Code tool whitelist
 - `--env KEY=VALUE`: MCP environment variable value (repeatable)

tests/test_agent_composition.py

@@ -1205,8 +1205,6 @@ def test_cursor_generates_subagent_and_mcp_json(self):
 
     # ── VS Code ────────────────────────────────────────────────
 
-    # ── Gemini CLI ─────────────────────────────────────────────
-
     # ── Kiro ───────────────────────────────────────────────────
 
     def test_kiro_generates_agent_json(self):

tests/test_agent_config_generator.py

@@ -415,12 +415,7 @@ def test_kiro_agent_profile_truncates_description(self):
 
 
 # ═══════════════════════════════════════════════════════════════════
-# 7. generate_agent_config — Gemini CLI
-# ═══════════════════════════════════════════════════════════════════
-
-
-# ═══════════════════════════════════════════════════════════════════
-# 8. generate_agent_config — Codex
+# 7. generate_agent_config — Codex
 # ═══════════════════════════════════════════════════════════════════
 
 
@@ -449,7 +444,7 @@ def test_content_not_empty(self):
 
 
 # ═══════════════════════════════════════════════════════════════════
-# 9. generate_agent_config — Copilot
+# 8. generate_agent_config — Copilot
 # ═══════════════════════════════════════════════════════════════════
 
 

tests/test_harness_config_e2e.py

@@ -6,7 +6,7 @@
 
 """Comprehensive end-to-end tests for issue #434: first-class harness support.
 
-Covers Codex CLI, Gemini CLI, GitHub Copilot (VS Code), and OpenCode across:
+Covers Codex CLI, GitHub Copilot (VS Code), and OpenCode across:
 
 1. Server-side config generation (agent_config_generator)
 2. CLI pull command (cmd_pull._dict_to_toml, _write_file, full pull flow)

tests/test_pull_and_agent_cli.py

@@ -281,12 +281,7 @@ def test_mcp_config_without_path_not_written(self, tmp_path: Path):
 
 
 # ═══════════════════════════════════════════════════════════════
-# 3. Gemini CLI format
-# ═══════════════════════════════════════════════════════════════
-
-
-# ═══════════════════════════════════════════════════════════════
-# 4. Kiro format (agent_profile with ~/  path)
+# 3. Kiro format (agent_profile with ~/  path)
 # ═══════════════════════════════════════════════════════════════