@ouro.bot/cli 0.1.0-alpha.3 → 0.1.0-alpha.300

package/skills/configure-dev-tools.md

@@ -0,0 +1,81 @@
+# Configure Dev Tools for MCP Agent Bridge
+
+Set up your development tools (Claude Code, Codex) to communicate with Ouroboros agents via MCP. One command does everything.
+
+## Setup
+
+### Claude Code
+
+```bash
+ouro setup --tool claude-code --agent <agent-name>
+```
+
+This command:
+1. Registers the MCP server with Claude Code via `claude mcp add`
+2. Configures lifecycle hooks (SessionStart, Stop, PostToolUse) for passive awareness
+3. Detects dev vs installed mode automatically and uses the correct command path
+
+### Codex
+
+```bash
+ouro setup --tool codex --agent <agent-name>
+```
+
+This command:
+1. Registers the MCP server with Codex via `codex mcp add`
+2. Detects dev vs installed mode automatically
+
+## Verification
+
+After setup, verify the connection:
+
+1. **Check daemon is running**: `ouro up` (installed) or `ouro dev --repo-path <path>` (dev mode).
+2. **Test from Claude Code**: Start a new session and use the `status` tool.
+3. **Test from Codex**: Run `codex exec "Use the <agent-name> status tool"`.
+4. **Check registration**: `claude mcp list` or `codex mcp list`.
+
+## Available MCP Tools
+
+Once connected, these tools are available:
+
+### Conversation tools (new)
+- **send_message** -- Send a message and get a synchronous agent response (full turn with tools)
+- **check_response** -- Check for pending messages from the agent (after ponder or proactive surface)
+
+### Read-only tools
+- **ask** -- Ask the agent a question (uses memory and context)
+- **status** -- Get agent's current status and activity
+- **catchup** -- Get recent activity summary
+- **get_context** -- Get agent's current working context
+- **search_memory** -- Search agent's memory for specific topics
+- **get_task** -- Get details of the agent's current task
+- **check_scope** -- Verify if something is in scope for current work
+- **check_guidance** -- Get guidance on how to approach something
+
+### Write tools
+- **delegate** -- Request the agent to handle a task (runs full conversation turn)
+- **request_decision** -- Ask agent to make a decision about something
+- **report_progress** -- Report progress on delegated work
+- **report_blocker** -- Report a blocker on delegated work
+- **report_complete** -- Report completion of delegated work
+
+## Troubleshooting
+
+### "Daemon not running" error
+Most read-only tools work without the daemon (reads filesystem directly). For write operations and `send_message`, start the daemon with `ouro up` or `ouro dev`.
+
+### MCP server not appearing
+- Run `claude mcp list` or `codex mcp list` to verify registration
+- Re-run `ouro setup` to fix
+- Restart your dev tool (MCP loads at session start)
+
+### Connection timeouts
+- Ensure `dist/` is built: `npm run build`
+- Check that the entry point path is correct (setup auto-detects this)
+
+### Removing
+```bash
+claude mcp remove ouro-<agent-name>
+# or
+codex mcp remove ouro-<agent-name>
+```

package/skills/travel-planning.md

@@ -0,0 +1,138 @@
+# Travel Planning Skill
+
+Compose all travel infrastructure to plan and book trips effectively.
+
+## Research Phase
+
+### Destination Research
+1. Use `geocode_search` to find candidate destinations and get coordinates
+2. Use `weather_lookup` to check climate for travel dates
+3. Use `travel_advisory` to check US State Dept safety levels (1-4):
+   - Level 1: Exercise Normal Precautions
+   - Level 2: Exercise Increased Caution
+   - Level 3: Reconsider Travel
+   - Level 4: Do Not Travel
+4. Use browser tools (see `browser-navigation` skill) for destination reviews, travel blogs, local tips
+5. Compare multiple destinations on weather, safety, cost, and flight availability
+
+### Information Gathering Workflow
+```
+geocode_search -> get coordinates
+weather_lookup -> check weather at coordinates
+travel_advisory -> check safety level
+browser tools -> reviews, local info, pricing
+```
+
+## Flight Search
+
+### Using Duffel MCP (when available)
+- Search via `duffel_search_flights` tool (first-class MCP tool)
+- Provide: origin, destination, dates, passengers, cabin class
+- Compare options on price, duration, stops, airline
+
+### Using Browser (fallback)
+- Use browser-navigation skill patterns for Google Flights, airline sites
+- Add delays between searches to avoid detection
+
+### Comparison Criteria
+Present top 3-5 options comparing:
+- Total price (including bags, seat selection)
+- Duration and number of stops
+- Airline and aircraft
+- Departure/arrival times
+- Cancellation/change policy
+
+## Accommodation Search
+
+### Hotel Search
+- Use Expedia MCP when available (`expedia_search_hotels` tool)
+- Compare: price per night, total stay cost, location, rating, amenities
+- Check cancellation policies (crucial for travel planning)
+
+### Vacation Rentals (Browser-based)
+- Use browser-navigation skill for Airbnb/VRBO
+- Search by location, dates, guests, budget
+- Compare: price, location proximity, reviews, amenities, host rating
+- Screenshot listings for user review
+
+### Comparison Criteria
+- Price per night and total cost
+- Location (distance to key attractions/activities)
+- Reviews and ratings
+- Amenities (WiFi, kitchen, parking, pool)
+- Cancellation flexibility
+
+## Booking Workflow
+
+### Critical Safety Gates
+- **ALWAYS confirm with user before any booking action**
+- **ALWAYS confirm before entering payment information**
+- **ALWAYS confirm before agreeing to terms**
+- **Never proceed with a financial commitment without explicit approval**
+
+### Credential Handling
+
+Credentials are managed through the credential access layer, which stores
+agent-owned secrets encrypted in the bundle vault. Raw passwords never
+enter model context.
+
+- Use `credential_get` to check what credentials exist for a domain (metadata only, never passwords)
+- Use `credential_store` to save credentials the agent acquired (e.g., during sign-up for a service)
+- The credential gateway automatically injects secrets into API requests via `getRawSecret()`
+- Use browser-navigation skill form patterns for entering credentials during interactive sessions
+
+**How credentials work:**
+- Agent-owned credentials are AES-256-GCM encrypted in `~/AgentBundles/<agent>.ouro/vault/`
+- The master key lives at `~/.agentsecrets/<agent>/vault.key` (auto-generated on first use)
+- The agent can sign up for services and store its own credentials
+- Stored passwords are never returned to the model — only metadata (domain, username, notes)
+
+**Upgrade path:** Install `aac` (Bitwarden Agent Access CLI) for vault-backed credentials.
+When `aac` is available, the harness can read credentials from Bitwarden directly,
+enabling access to human-managed vault items without storing them locally.
+
+### Post-Booking
+- Save confirmation details (confirmation number, dates, hotel name, airline, booking reference)
+- Save to diary/journal for future reference
+- Set reminders for check-in windows
+- Note cancellation deadlines
+
+## Preference Management
+
+### Storing Preferences
+Track and reference these travel preferences:
+- Preferred airlines and alliance status
+- Preferred hotel chains and loyalty numbers
+- Seat preferences (window/aisle, extra legroom)
+- Dietary needs for in-flight meals
+- Budget ranges (per night for hotels, per flight)
+- Must-have amenities (WiFi, gym, pool)
+
+### Using Preferences
+- Reference stored preferences when searching
+- Apply airline preferences to flight comparisons
+- Apply hotel brand preferences to accommodation searches
+- Adjust budget ranges based on destination
+
+## Tool Reference
+
+### Native Tools (always available)
+- `weather_lookup` - Current weather and daily forecast by city or coordinates (Open-Meteo, zero config)
+- `travel_advisory` - US State Dept advisory by country code
+- `geocode_search` - Location/POI search with coordinates
+- `credential_get` - Check credential metadata for a domain (never returns passwords)
+- `credential_store` - Store credentials the agent acquired (family trust, confirmation required)
+- `credential_list` - List stored credential domains
+- `credential_delete` - Delete stored credentials (family trust, confirmation required)
+
+### MCP Tools (when configured)
+- Browser tools via `@playwright/mcp` - see `browser-navigation` skill
+- Duffel flight search (when MCP server available)
+- Expedia hotel search (when MCP server available)
+
+### Human Confirmation Required For
+- Any booking or payment
+- Entering personal information
+- Agreeing to terms of service
+- Creating financial obligations
+- Sharing credentials with third parties

package/AdoptionSpecialist.ouro/agent.json

@@ -1,20 +0,0 @@
-{
-  "version": 1,
-  "enabled": false,
-  "provider": "anthropic",
-  "context": {
-    "maxTokens": 80000,
-    "contextMargin": 20
-  },
-  "phrases": {
-    "thinking": [
-      "matching hatchlings"
-    ],
-    "tool": [
-      "checking adoption notes"
-    ],
-    "followup": [
-      "finalizing hatch plan"
-    ]
-  }
-}

package/AdoptionSpecialist.ouro/psyche/identities/python.md

@@ -1 +0,0 @@
-# Python

package/dist/heart/daemon/subagent-installer.js

@@ -1,125 +0,0 @@
-"use strict";
-var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
-    if (k2 === undefined) k2 = k;
-    var desc = Object.getOwnPropertyDescriptor(m, k);
-    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
-      desc = { enumerable: true, get: function() { return m[k]; } };
-    }
-    Object.defineProperty(o, k2, desc);
-}) : (function(o, m, k, k2) {
-    if (k2 === undefined) k2 = k;
-    o[k2] = m[k];
-}));
-var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
-    Object.defineProperty(o, "default", { enumerable: true, value: v });
-}) : function(o, v) {
-    o["default"] = v;
-});
-var __importStar = (this && this.__importStar) || (function () {
-    var ownKeys = function(o) {
-        ownKeys = Object.getOwnPropertyNames || function (o) {
-            var ar = [];
-            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
-            return ar;
-        };
-        return ownKeys(o);
-    };
-    return function (mod) {
-        if (mod && mod.__esModule) return mod;
-        var result = {};
-        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
-        __setModuleDefault(result, mod);
-        return result;
-    };
-})();
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.installSubagentsForAvailableCli = installSubagentsForAvailableCli;
-const fs = __importStar(require("fs"));
-const os = __importStar(require("os"));
-const path = __importStar(require("path"));
-const child_process_1 = require("child_process");
-const runtime_1 = require("../../nerves/runtime");
-function detectCliBinary(binary) {
-    const result = (0, child_process_1.spawnSync)("which", [binary], { encoding: "utf-8" });
-    if (result.status !== 0)
-        return null;
-    const resolved = result.stdout.trim();
-    return resolved.length > 0 ? resolved : null;
-}
-function listSubagentSources(subagentsDir) {
-    if (!fs.existsSync(subagentsDir))
-        return [];
-    return fs.readdirSync(subagentsDir)
-        .filter((name) => name.endsWith(".md"))
-        .filter((name) => name !== "README.md")
-        .map((name) => path.join(subagentsDir, name))
-        .sort((a, b) => a.localeCompare(b));
-}
-function ensureSymlink(source, target) {
-    fs.mkdirSync(path.dirname(target), { recursive: true });
-    if (fs.existsSync(target)) {
-        const stats = fs.lstatSync(target);
-        if (stats.isSymbolicLink()) {
-            const linkedPath = fs.readlinkSync(target);
-            if (linkedPath === source)
-                return false;
-        }
-        fs.unlinkSync(target);
-    }
-    fs.symlinkSync(source, target);
-    return true;
-}
-async function installSubagentsForAvailableCli(options = {}) {
-    const repoRoot = options.repoRoot ?? path.resolve(__dirname, "..", "..", "..");
-    const homeDir = options.homeDir ?? os.homedir();
-    const which = options.which ?? detectCliBinary;
-    const subagentsDir = path.join(repoRoot, "subagents");
-    const sources = listSubagentSources(subagentsDir);
-    const notes = [];
-    (0, runtime_1.emitNervesEvent)({
-        component: "daemon",
-        event: "daemon.subagent_install_start",
-        message: "starting subagent auto-install",
-        meta: { sources: sources.length },
-    });
-    if (sources.length === 0) {
-        notes.push(`no subagent files found at ${subagentsDir}`);
-        return { claudeInstalled: 0, codexInstalled: 0, notes };
-    }
-    let claudeInstalled = 0;
-    let codexInstalled = 0;
-    const claudePath = which("claude");
-    if (!claudePath) {
-        notes.push("claude CLI not found; skipping subagent install");
-    }
-    else {
-        const claudeAgentsDir = path.join(homeDir, ".claude", "agents");
-        for (const source of sources) {
-            const target = path.join(claudeAgentsDir, path.basename(source));
-            if (ensureSymlink(source, target)) {
-                claudeInstalled += 1;
-            }
-        }
-    }
-    const codexPath = which("codex");
-    if (!codexPath) {
-        notes.push("codex CLI not found; skipping subagent install");
-    }
-    else {
-        const codexSkillsDir = path.join(homeDir, ".codex", "skills");
-        for (const source of sources) {
-            const skillName = path.basename(source, ".md");
-            const target = path.join(codexSkillsDir, skillName, "SKILL.md");
-            if (ensureSymlink(source, target)) {
-                codexInstalled += 1;
-            }
-        }
-    }
-    (0, runtime_1.emitNervesEvent)({
-        component: "daemon",
-        event: "daemon.subagent_install_end",
-        message: "completed subagent auto-install",
-        meta: { claudeInstalled, codexInstalled, notes: notes.length },
-    });
-    return { claudeInstalled, codexInstalled, notes };
-}

package/dist/inner-worker-entry.js

@@ -1,4 +0,0 @@
-"use strict";
-Object.defineProperty(exports, "__esModule", { value: true });
-// Backward-compatible wrapper; unified runtime now lives at heart/agent-entry.
-require("./heart/agent-entry");

package/subagents/README.md

@@ -1,73 +0,0 @@
-# Sub-agents
-
-These are source-of-truth workflow definitions (`work-planner`, `work-doer`, `work-merger`) for planning, execution, and merge. They can be consumed either as Claude sub-agents (`.md` files with YAML frontmatter) or as Codex-style skills (`SKILL.md`).
-
-## Installation
-
-### Claude Code (sub-agents)
-
-Copy or symlink these files into Claude's sub-agent directory:
-
-```bash
-# Claude Code
-cp subagents/*.md ~/.claude/agents/
-# or
-ln -s "$(pwd)"/subagents/*.md ~/.claude/agents/
-```
-
-### Codex / skill-based harnesses
-
-For tools that support skills but not Claude sub-agents, install these as skills:
-
-```bash
-mkdir -p ~/.codex/skills/work-planner ~/.codex/skills/work-doer ~/.codex/skills/work-merger
-
-# Hard-link to keep one source of truth
-ln -f "$(pwd)/subagents/work-planner.md" ~/.codex/skills/work-planner/SKILL.md
-ln -f "$(pwd)/subagents/work-doer.md" ~/.codex/skills/work-doer/SKILL.md
-ln -f "$(pwd)/subagents/work-merger.md" ~/.codex/skills/work-merger/SKILL.md
-```
-
-**Important:** Hard-links break when editors save by replacing the file (new inode). After editing any `subagents/*.md` file, re-run the `ln -f` command for that file to restore the link. You can verify with `stat -f '%i'` — both files should share the same inode.
-
-Optional UI metadata:
-
-```bash
-mkdir -p ~/.codex/skills/work-planner/agents ~/.codex/skills/work-doer/agents ~/.codex/skills/work-merger/agents
-cat > ~/.codex/skills/work-planner/agents/openai.yaml << 'EOF'
-interface:
-  display_name: "Work Planner"
-  short_description: "Create and gate planning/doing task docs"
-  default_prompt: "Use $work-planner to create or update a planning doc, then stop at NEEDS_REVIEW."
-EOF
-cat > ~/.codex/skills/work-doer/agents/openai.yaml << 'EOF'
-interface:
-  display_name: "Work Doer"
-  short_description: "Execute approved doing docs with strict TDD"
-  default_prompt: "Use $work-doer to execute an approved doing doc unit by unit."
-EOF
-cat > ~/.codex/skills/work-merger/agents/openai.yaml << 'EOF'
-interface:
-  display_name: "Work Merger"
-  short_description: "Merge feature branch into main via PR after work-doer completes"
-  default_prompt: "Use $work-merger to merge the current feature branch into main."
-EOF
-```
-
-Restart the harness after install so new skills are discovered.
-
-## Available sub-agents
-
-| File | Purpose |
-|------|---------|
-| `work-planner.md` | Interactive task planner. Generates planning docs through conversation, then converts to doing docs after human approval. |
-| `work-doer.md` | Task executor. Reads a doing doc and works through each unit sequentially with strict TDD. |
-| `work-merger.md` | Sync-and-merge agent. Merges feature branch into main via PR after work-doer completes. Handles conflicts, CI failures, and race conditions. |
-
-## Workflow
-
-1. Human describes a task
-2. Agent invokes **work-planner** to create a planning doc → human approves → planner converts to doing doc
-3. Agent invokes **work-doer** to execute the doing doc unit by unit
-4. Each unit is committed independently with progress tracked in the doing doc
-5. Agent invokes **work-merger** to merge the feature branch into main via PR (fetch, merge, resolve conflicts, CI gate, merge PR, cleanup)

package/subagents/work-doer.md

@@ -1,233 +0,0 @@
----
-name: work-doer
-description: Executes doing.md units sequentially with strict TDD. Reads the doing doc, works through each unit, commits after each. Use after planning is complete and doing.md exists.
-model: opus
----
-
-You are a task executor. Read a doing.md file and execute all units sequentially until complete or blocked.
-
-## On Startup
-
-1. **Find doing doc**: Look for `YYYY-MM-DD-HHMM-doing-*.md` in repo root
-2. If multiple found, ask which one
-3. If none found, ask user for location
-4. **Check execution_mode**: Read the doing doc's `Execution Mode` field
-5. **Verify artifacts directory exists**: `{task-name}/` next to `{task-name}.md`
-   - If missing, create it: `mkdir {task-name}`
-6. **Detect resume vs fresh start:**
-   - Count completed units (✅) vs total units
-   - Check git status for uncommitted changes
-
-7. **Announce status clearly:**
-
-**If fresh start (0 units complete):**
-```
-found: YYYY-MM-DD-HHMM-doing-{name}.md
-execution_mode: [pending|spawn|direct]
-artifacts: ./{task-name}/
-status: fresh start
-units: 0/X complete
-starting Unit 0...
-```
-
-**If resuming (some units complete):**
-```
-found: YYYY-MM-DD-HHMM-doing-{name}.md
-execution_mode: [pending|spawn|direct]
-status: RESUMING
-units: Y/X complete (✅ Unit 0, 1a, 1b...)
-uncommitted changes: [yes/no]
-resuming from Unit Z...
-```
-
-**If uncommitted changes detected:**
-```
-⚠️ uncommitted changes found
-recommend: commit or stash before continuing
-proceed anyway? (y/n)
-```
-
----
-
-## Timestamp & Commit Pattern
-
-**All timestamps come from git commits for audit trail.**
-
-To get timestamp for progress log entries:
-```bash
-git log -1 --format="%Y-%m-%d %H:%M"
-```
-
-After any edit to doing doc:
-1. Stage: `git add doing-*.md`
-2. Commit: `git commit -m "docs(doing): <what changed>"`
-3. Get timestamp from git log
-4. Use that timestamp in progress log entry
-
----
-
-## Execution Loop
-
-For each unit in order:
-
-### 1. Announce
-```
-starting Unit Xa: [name]
-```
-
-### 2. Execute (TDD strictly enforced)
-
-**General execution rules:**
-- Save all outputs, logs, and data to `{task-name}/` artifacts directory
-- If execution_mode is `pending`, wait for user approval before starting each unit
-- If execution_mode is `spawn`, spawn a sub-agent for each unit
-- If execution_mode is `direct`, proceed immediately
-
-**For test units (Xa):**
-1. Write failing tests for the feature
-2. Run tests — **must FAIL (red)**
-3. If tests pass immediately, something is wrong — investigate
-4. Commit: `git commit -m "test(scope): Unit Xa - [description]"`
-5. Push
-
-**For implementation units (Xb):**
-1. Write minimal code to make tests pass
-2. **Do NOT modify tests** — implementation must satisfy existing tests
-3. Run tests — **must PASS (green)**
-4. **Run the build** (e.g. `npm run build`, `cargo build`, `go build`) — the project must compile with no errors. Tests alone are not sufficient (test runners may handle imports/modules differently than the real compiler).
-5. No warnings allowed
-6. Commit: `git commit -m "feat(scope): Unit Xb - [description]"`
-7. Push
-
-**For verify/refactor units (Xc):**
-1. Run coverage report
-2. **Must be 100% on new code** — if not, add tests
-3. Check edge cases: null, empty, boundary values
-4. Check all error paths tested
-5. Refactor if needed, keep tests green
-6. **Run the build** — verify the project compiles clean
-7. Commit: `git commit -m "refactor(scope): Unit Xc - [description]"` (if changes made)
-8. Push
-
-**For non-coding units:**
-1. Complete work as described
-2. Produce specified output
-3. Verify acceptance criteria
-4. Commit relevant files
-5. Push
-
-### 3. Update doing.md
-- Change unit status: `⬜` → `✅`
-- Update `Completion Criteria` checkboxes that are now satisfied by this unit's evidence
-- Commit: `git commit -m "docs(doing): complete Unit Xa"`
-- Get timestamp: `git log -1 --format="%Y-%m-%d %H:%M"`
-- Add progress log entry with that timestamp:
-  ```
-  - 2026-02-03 14:25 Unit Xa complete: [brief summary]
-  ```
-
-### 4. Context management
-- Run `/compact` between units if context growing large
-- Each unit should be independent
-- Re-read files if you need prior context
-
-### 5. Continue to next unit
-
----
-
-## Code Coverage Requirements
-
-**MANDATORY: 100% coverage on all new code.**
-
-Before marking any implementation unit complete:
-1. Run coverage report
-2. Verify 100% on new/modified files
-3. No `[ExcludeFromCodeCoverage]` or equivalent on new code
-4. All branches covered (if/else, switch, try/catch)
-5. All error paths have tests
-6. If coverage < 100%, add tests before proceeding
-
----
-
-## TDD Requirements
-
-**Strict TDD — no exceptions:**
-
-1. **Tests first**: Write failing tests BEFORE any implementation
-2. **Red**: Run tests, confirm they FAIL
-3. **Green**: Write minimal code to pass
-4. **Refactor**: Clean up, tests stay green
-5. **Never skip**: No implementation without failing test first
-6. **Never modify tests to pass**: Implementation satisfies tests, not vice versa
-
----
-
-## Blocker Handling
-
-**For simple fixes or test failures:**
-1. **Spawn sub-agent immediately** — don't ask, just do it
-2. Sub-agent analyzes error, fixes issue, commits, pushes
-3. Sub-agent reports back when done
-4. Continue with next unit
-
-**For actual blockers (requirements unclear, external dependency, design decision needed):**
-1. Mark unit as `❌ Blocked` in doing.md
-2. Commit: `git commit -m "docs(doing): Unit Xa blocked"`
-3. Get timestamp from git
-4. Add progress log entry with error details
-5. Output:
-   ```
-   ❌ blocked on Unit Xa
-   error: [description]
-   tried: [what you attempted]
-   need: [what would help]
-   ```
-6. **STOP** — do not proceed until user resolves
-
-**Rule of thumb:**
-- Code error / test failure → spawn sub-agent
-- Requirement unclear / need user input → mark blocked and stop
-
----
-
-## Completion
-
-When all units are `✅`:
-1. Run full test suite one final time
-2. Verify all tests pass, no warnings
-3. Mark all satisfied `Completion Criteria` checkboxes in doing doc as `[x]`
-4. If `Planning:` doc path exists, sync its `Completion Criteria` checkboxes to `[x]` based on final evidence
-5. Update doing.md Status to `done`
-6. Commit: `git commit -m "docs(doing): all units complete"`
-7. Get timestamp from git
-8. Add final progress log entry
-9. Output:
-   ```
-   ✅ all units complete
-   tests: [X passing]
-   coverage: [X%]
-   status: done
-   ```
-
----
-
-## Rules
-
-1. **File naming**: Expect `YYYY-MM-DD-HHMM-doing-{name}.md` format
-2. **Artifacts directory**: Use `{task-name}/` for all outputs, logs, data
-3. **Execution mode**: Honor `pending | spawn | direct` from doing doc
-4. **TDD strictly enforced** — tests before implementation, always
-5. **100% coverage** — no exceptions, no exclude attributes
-6. **Atomic commits** — one logical unit per commit, push after each
-7. **Timestamps from git** — `git log -1 --format="%Y-%m-%d %H:%M"`
-8. **Push after each unit phase complete**
-9. **Update doing.md after each unit** — status and progress log
-10. **Spawn sub-agents for fixes** — don't ask, just do it
-11. **Update docs immediately** — when decisions made, commit right away
-12. **Stop on actual blocker** — unclear requirements or need user input
-13. **/compact proactively** — preserve context between units
-14. **No warnings** — treat warnings as errors
-15. **Run full test suite** — before marking unit complete, not just new tests
-16. **Always compile** — run the project's build command after every implementation/refactor unit. Tests passing is necessary but not sufficient.
-17. **Checklist hygiene is mandatory** — keep doing/planning `Completion Criteria` checklists synchronized with verified completion evidence.
-18. **Verify APIs before importing** — before writing `import { Foo } from './bar'`, use `grep` or `read_file` to confirm `Foo` is actually exported from that module. Never assume an export exists — always check the source first. This prevents wasted cycles on "module has no exported member" errors.