@pasko70/pibo 1.9.13 → 1.10.1

package/skills/builtin/ralph-prd-json/SKILL.md

@@ -1,123 +1,123 @@
----
-name: ralph-prd-json
-description: Convert Markdown PRDs and feature specs into Ralph prd.json story batches. Use whenever the user asks to convert a PRD for Ralph, create Ralph JSON, split a PRD into executable stories, prepare PRD story files, or mentions prd.json for autonomous implementation.
----
-
-# Ralph PRD JSON Converter
-
-Convert existing PRDs, specs, or Markdown feature descriptions into the `prd.json` format Ralph can execute.
-
-Use this skill for **PRD/story shaping**, not for creating or operating the Ralph job itself. For job creation, Docker worker reuse, worktrees, stop policies, and monitoring, use the `ralph-loop` skill.
-
-## Output Format
-
-```json
-{
-  "project": "[Project Name]",
-  "branchName": "ralph/[feature-name-kebab-case]",
-  "description": "[Feature description from PRD title/intro]",
-  "userStories": [
-    {
-      "id": "US-001",
-      "title": "[Story title]",
-      "description": "As a [user], I want [feature] so that [benefit]",
-      "acceptanceCriteria": [
-        "Criterion 1",
-        "Criterion 2",
-        "Typecheck passes"
-      ],
-      "priority": 1,
-      "passes": false,
-      "notes": ""
-    }
-  ]
-}
-```
-
-## Story Size Rule
-
-Each story should fit in one Ralph iteration and one context window. Split broad requirements until each story has a focused implementation and verification path.
-
-Good story sizes:
-
-- Add a database column and migration.
-- Add one CLI subcommand and tests.
-- Add one service method plus focused tests.
-- Add one UI component to an existing page.
-
-Too large:
-
-- Build the whole dashboard.
-- Add the entire telemetry system.
-- Refactor the API.
-
-Rule of thumb: if the change cannot be described in 2 to 3 sentences, split it.
-
-## Story Ordering
-
-Order by dependencies:
-
-1. Schema and migrations.
-2. Store/service layer.
-3. Runtime instrumentation or backend logic.
-4. CLI/API/UI surfaces.
-5. Docs and validation.
-
-Earlier stories must not depend on later stories.
-
-## Acceptance Criteria
-
-Use criteria Ralph can verify.
-
-Good:
-
-- `Add telemetry_turns table with session, turn, status, timestamps, and retention fields`.
-- `pibo debug telemetry stats --json returns counts by retention class`.
-- `Focused tests pass`.
-- `Typecheck passes`.
-
-Bad:
-
-- `Works correctly`.
-- `Good UX`.
-- `Handles everything`.
-
-Always include `Typecheck passes`. Add `Tests pass` for logic. Add browser verification when a browser UI is affected, and PTY or interactive verification when a CLI/TUI flow is affected.
-
-### Real-path criteria for user-facing work
-
-For CLI, TUI, Web UI, gateway, auth, runtime, persistence, or agent-routing stories, include an acceptance criterion for the closest practical default user path. Keep it proportional: a small component may only need a render test, while a messaging or navigation feature should exercise the real command, route, API, or runtime path when feasible.
-
-Fake sources, demo mode, mocks, and snapshots are good supporting evidence, but avoid making them the only evidence for behavior users will exercise directly. If the real path cannot be tested in the story, add an explicit criterion or note explaining why and what later story will cover it.
-
-Name the observable result, not just the mechanism. Prefer criteria like:
-
-- `Running pibo tui:sessions without --demo can create/select a session, send a message through the real router, and show an assistant reply`.
-- `The agent picker lists custom agents from the canonical custom-agent store, including an existing pibo-agent fixture or documented local equivalent`.
-- `Opening /sessions/<id> in the worker Web UI shows the session created by the CLI flow`.
-
-### Evidence notes
-
-When a story is marked `passes: true`, notes should record the concrete evidence: commands run, whether the path was fake/demo/real, browser or PTY checks performed, and any manual observation. `Tests pass` alone is not enough for user-facing integration stories.
-
-## Conversion Rules
-
-1. One user story per JSON entry.
-2. IDs are sequential, e.g. `US-001`.
-3. Priority follows dependency order, then document order.
-4. New stories start with `passes: false` and empty `notes`.
-5. `branchName` is kebab-case and prefixed with `ralph/`.
-6. Keep stories independent enough that Ralph can commit after each story or coherent group.
-
-## Checklist Before Saving
-
-- [ ] Stories fit one iteration.
-- [ ] Stories are ordered by dependency.
-- [ ] Acceptance criteria are verifiable.
-- [ ] Every story has `Typecheck passes`.
-- [ ] Logic stories include `Tests pass`.
-- [ ] UI stories include browser verification when a browser surface is affected.
-- [ ] CLI/TUI stories include PTY or interactive validation when the interactive path is in scope and feasible.
-- [ ] User-facing integration stories include a real/default-path check or an explicit reason it is deferred.
-- [ ] Fake/demo/mocked checks are not the only evidence for default-path behavior unless documented as a deliberate limitation.
-- [ ] No story depends on a later story.
+---
+name: ralph-prd-json
+description: Convert Markdown PRDs and feature specs into Ralph prd.json story batches. Use whenever the user asks to convert a PRD for Ralph, create Ralph JSON, split a PRD into executable stories, prepare PRD story files, or mentions prd.json for autonomous implementation.
+---
+
+# Ralph PRD JSON Converter
+
+Convert existing PRDs, specs, or Markdown feature descriptions into the `prd.json` format Ralph can execute.
+
+Use this skill for **PRD/story shaping**, not for creating or operating the Ralph job itself. For job creation, Docker worker reuse, worktrees, stop policies, and monitoring, use the `ralph-loop` skill.
+
+## Output Format
+
+```json
+{
+  "project": "[Project Name]",
+  "branchName": "ralph/[feature-name-kebab-case]",
+  "description": "[Feature description from PRD title/intro]",
+  "userStories": [
+    {
+      "id": "US-001",
+      "title": "[Story title]",
+      "description": "As a [user], I want [feature] so that [benefit]",
+      "acceptanceCriteria": [
+        "Criterion 1",
+        "Criterion 2",
+        "Typecheck passes"
+      ],
+      "priority": 1,
+      "passes": false,
+      "notes": ""
+    }
+  ]
+}
+```
+
+## Story Size Rule
+
+Each story should fit in one Ralph iteration and one context window. Split broad requirements until each story has a focused implementation and verification path.
+
+Good story sizes:
+
+- Add a database column and migration.
+- Add one CLI subcommand and tests.
+- Add one service method plus focused tests.
+- Add one UI component to an existing page.
+
+Too large:
+
+- Build the whole dashboard.
+- Add the entire telemetry system.
+- Refactor the API.
+
+Rule of thumb: if the change cannot be described in 2 to 3 sentences, split it.
+
+## Story Ordering
+
+Order by dependencies:
+
+1. Schema and migrations.
+2. Store/service layer.
+3. Runtime instrumentation or backend logic.
+4. CLI/API/UI surfaces.
+5. Docs and validation.
+
+Earlier stories must not depend on later stories.
+
+## Acceptance Criteria
+
+Use criteria Ralph can verify.
+
+Good:
+
+- `Add telemetry_turns table with session, turn, status, timestamps, and retention fields`.
+- `pibo debug telemetry stats --json returns counts by retention class`.
+- `Focused tests pass`.
+- `Typecheck passes`.
+
+Bad:
+
+- `Works correctly`.
+- `Good UX`.
+- `Handles everything`.
+
+Always include `Typecheck passes`. Add `Tests pass` for logic. Add browser verification when a browser UI is affected, and PTY or interactive verification when a CLI/TUI flow is affected.
+
+### Real-path criteria for user-facing work
+
+For CLI, TUI, Web UI, gateway, auth, runtime, persistence, or agent-routing stories, include an acceptance criterion for the closest practical default user path. Keep it proportional: a small component may only need a render test, while a messaging or navigation feature should exercise the real command, route, API, or runtime path when feasible.
+
+Fake sources, demo mode, mocks, and snapshots are good supporting evidence, but avoid making them the only evidence for behavior users will exercise directly. If the real path cannot be tested in the story, add an explicit criterion or note explaining why and what later story will cover it.
+
+Name the observable result, not just the mechanism. Prefer criteria like:
+
+- `Running pibo tui:sessions without --demo can create/select a session, send a message through the real router, and show an assistant reply`.
+- `The agent picker lists custom agents from the canonical custom-agent store, including an existing pibo-agent fixture or documented local equivalent`.
+- `Opening /sessions/<id> in the worker Web UI shows the session created by the CLI flow`.
+
+### Evidence notes
+
+When a story is marked `passes: true`, notes should record the concrete evidence: commands run, whether the path was fake/demo/real, browser or PTY checks performed, and any manual observation. `Tests pass` alone is not enough for user-facing integration stories.
+
+## Conversion Rules
+
+1. One user story per JSON entry.
+2. IDs are sequential, e.g. `US-001`.
+3. Priority follows dependency order, then document order.
+4. New stories start with `passes: false` and empty `notes`.
+5. `branchName` is kebab-case and prefixed with `ralph/`.
+6. Keep stories independent enough that Ralph can commit after each story or coherent group.
+
+## Checklist Before Saving
+
+- [ ] Stories fit one iteration.
+- [ ] Stories are ordered by dependency.
+- [ ] Acceptance criteria are verifiable.
+- [ ] Every story has `Typecheck passes`.
+- [ ] Logic stories include `Tests pass`.
+- [ ] UI stories include browser verification when a browser surface is affected.
+- [ ] CLI/TUI stories include PTY or interactive validation when the interactive path is in scope and feasible.
+- [ ] User-facing integration stories include a real/default-path check or an explicit reason it is deferred.
+- [ ] Fake/demo/mocked checks are not the only evidence for default-path behavior unless documented as a deliberate limitation.
+- [ ] No story depends on a later story.