@synergenius/flow-weaver-pack-weaver 0.9.185 → 0.9.187

package/src/bot/capability-registry.ts

@@ -48,6 +48,21 @@ Your only output is task_create calls + done.
 6. Maximize parallelism: tasks with no shared files should not depend on each other.
 7. Your LAST subtask: "Review & Steer" assigned to orchestrator, dependsOn all others.
 
+### Design First
+For multi-file projects, your FIRST subtask should create a .design.md with:
+- Module map: which files, what each exports (function signatures with types)
+- Conventions: naming, error handling, patterns (e.g., factory functions for servers)
+All subsequent developer tasks must read .design.md before writing code.
+
+### Acceptance Criteria
+Every subtask MUST have acceptance.checks — shell commands that verify "done" (exit 0 = pass).
+The system runs them automatically after each run. If any fail, the task stays open.
+Examples:
+- File exists: test -f url-shortener/src/store.ts
+- Compiles: cd url-shortener && npx tsc --noEmit
+- Tests pass: cd url-shortener && npx vitest run
+- Export exists: grep -q "export.*startServer" url-shortener/src/server.ts
+
 ### Steering Mode (when running a Review & Steer task)
 Read sibling task statuses and acceptance results from your context, then decide:
 - All checks pass → done.

package/dist/docs/weaver-bot-usage.md

@@ -1,51 +0,0 @@
-## Weaver Bot (Swarm)
-
-The Weaver bot is an autonomous AI agent that creates and modifies Flow Weaver workflows. Tasks are executed via the swarm orchestrator, which routes work to the best available bot profile.
-
-### Running tasks
-
-Use `fw_weaver_task_create` to submit work:
-- `title`: Short task title (required)
-- `description`: Detailed instructions
-- `assignedProfile`: Target a specific bot profile
-- `complexity`: trivial, simple, moderate, complex
-- `subtasks`: Inline subtask definitions with dependency chaining
-
-Then start the swarm with `fw_weaver_swarm_start` to begin execution.
-
-### Execution flow
-
-1. **Create task** — `fw_weaver_task_create` adds task to the pool
-2. **Orchestrator routes** — matches task to best bot profile via fast-path cascade (exact-match, capability-match, single-eligible, ai-routed, round-robin)
-3. **Build context** — gathers project state, templates, and relevant files
-4. **Plan** — AI generates a step-by-step execution plan
-5. **Execute + validate + retry** — runs steps, validates output, retries on errors (up to maxAttempts)
-6. **Git ops** — commits changes if successful
-7. **Report** — task marked done/failed with run summary
-
-### Steering a running bot
-
-Use `fw_weaver_steer` with a `botId` to control execution:
-- `pause` — pause at next safe point
-- `resume` — continue after pause
-- `cancel` — abort execution
-
-### Checking status
-
-- `fw_weaver_status` — quick swarm summary (active bots, tasks completed, budget)
-- `fw_weaver_swarm_status` — full swarm state with per-instance details
-- `fw_weaver_swarm_events` — live event stream
-- `fw_weaver_orchestrator_status` — routing decisions, stats, active instances
-
-### Profiles
-
-Use `fw_weaver_profile_create` to create specialized bot profiles. Each profile defines:
-
-- **capabilities** — free-form list of `{ name, description }` objects describing what the profile can do. Names are arbitrary strings (e.g. `code-generation`, `testing`, `deployment`). The orchestrator uses the description text to match tasks to profiles.
-- **preferences** — intent-based settings that control profile behavior:
-  - `costStrategy` — `"economy"` (minimize cost), `"balanced"` (default), or `"quality"` (maximize quality regardless of cost)
-  - `requireApproval` — whether tasks routed to this profile need human approval before execution
-  - `instructions` — optional free-text instructions appended to the bot's system prompt
-- **scaling** — `minInstances` and `maxInstances` control how many concurrent bots can run this profile
-
-The orchestrator routes tasks to profiles based on `assignedProfile` or by matching task intent against profile capabilities.

package/dist/docs/weaver-genesis.md

@@ -1,32 +0,0 @@
-## Genesis Self-Evolution
-
-Genesis is Weaver's self-evolution protocol. It autonomously observes, proposes, applies, and validates changes to workflows.
-
-### Running a Genesis cycle
-
-Use `fw_weaver_genesis` to trigger a single cycle:
-- `projectDir`: Project directory (defaults to workspace)
-- `dryRun`: Preview changes without applying
-
-### How it works
-
-1. **Observe** — scans project for workflows and their health
-2. **Fingerprint** — detects if workflows changed since last cycle (avoids duplicate proposals)
-3. **Stabilize check** — prevents rapid churn (cooldown between cycles)
-4. **Propose** — AI analyzes project state and proposes modifications within a budget
-5. **Validate proposal** — checks proposed changes are safe
-6. **Snapshot** — creates backup before applying
-7. **Apply** — executes proposed operations via `flow-weaver modify`
-8. **Validate result** — compiles and validates the modified workflow
-9. **Threshold check** — decides if changes meet the quality bar
-10. **Approve** — human approval gate (configurable)
-11. **Commit** — git commit with rollback capability
-12. **Escrow** — data safety pipeline for recovery
-
-### Safety features
-
-- Pre-apply snapshots for rollback
-- Escrow recovery from previous cycles
-- Fingerprinting prevents duplicate proposals
-- Threshold checking before committing
-- Budget limits on proposed changes

package/dist/docs/weaver-task-queue.md

@@ -1,46 +0,0 @@
-## Task Management & Steering
-
-Weaver uses a swarm-based task system for background processing and real-time bot control.
-
-### Creating tasks
-
-Use `fw_weaver_task_create` to create tasks:
-- `title` — short task title (required)
-- `description` — detailed instructions
-- `priority` — higher number = picked first (default 0)
-- `assignedProfile` — route to a specific bot profile
-- `complexity` — trivial, simple, moderate, complex
-- `subtasks` — inline subtask definitions with `^prev` dependency shorthand
-- `budgetTokens` / `budgetCost` — per-task budget limits
-
-### Listing and managing tasks
-
-- `fw_weaver_task_list` — list tasks (filter by status, parentId, botId)
-- `fw_weaver_task_get` — get task details including subtasks
-- `fw_weaver_task_update` — update task fields
-- `fw_weaver_task_cancel` — cancel a pending or running task
-- `fw_weaver_task_retry` — reset a failed task back to pending
-
-### Steering commands
-
-Use `fw_weaver_steer` to control a running bot instance:
-- `pause` — pause execution at the next safe point
-- `resume` — continue after a pause
-- `cancel` — abort the current task
-
-### Swarm control
-
-- `fw_weaver_swarm_start` — start the swarm (spawns bot loops that claim and execute tasks)
-- `fw_weaver_swarm_pause` — pause all bot loops
-- `fw_weaver_swarm_stop` — stop the swarm gracefully
-- `fw_weaver_swarm_status` — full swarm state with per-bot details
-- `fw_weaver_swarm_config` — configure max concurrency, budgets, auto-retry
-- `fw_weaver_swarm_events` — stream swarm-level events for live updates
-
-### Monitoring
-
-Use `fw_weaver_status` to check the current swarm state:
-- Swarm phase (idle, running, paused, stopping)
-- Active/total bot instances
-- Tasks completed/failed
-- Budget usage