Menu
Run the full story lifecycle with delegation gates. Use to orchestrate a story.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Interactive adapter setup for Jira and Confluence. Detects available backends, discovers projects/spaces, generates validated YAML config. 3-4 questions max.
Evaluate design proportionality using Beck's four rules. Use after implementation.
Root cause analysis using the method best suited to the bug. Phase 3 of bugfix pipeline.
Push branch, create MR, verify artifacts complete. Phase 7 of bugfix pipeline.
Execute fix tasks with TDD and all validation gates. Phase 5 of bugfix pipeline.
Decompose fix into atomic TDD tasks. Phase 4 of bugfix pipeline.
| name | rai-story-run |
| description | Run the full story lifecycle with delegation gates. Use to orchestrate a story. |
| allowed-tools | ["Read","Grep","Glob","Bash","Agent","Skill"] |
| license | MIT |
| metadata | {"raise.work_cycle":"story","raise.frequency":"per-story","raise.fase":"","raise.prerequisites":"","raise.next":"","raise.gate":"","raise.adaptable":"true","raise.version":"2.1.0","raise.visibility":"public","raise.inputs":"- story_id: string, required, argument (e.g. \"S325.6\")\n- epic_id: string, optional, argument (inferred from story_id prefix)\n","raise.outputs":"- merge_commit: string, git\n- retrospective_md: file_path (work/epics/.../stories/s{N}.{M}-retrospective.md)\n- patterns: list, cli (via rai pattern add)\n"} |
Execute the full story lifecycle in one invocation, pausing only at delegation gates and resuming automatically from the last completed phase.
When to use: Starting or resuming any story. Replaces manual sequential skill invocation.
When to skip: Single-phase work (e.g., only running review on an already-implemented story). Individual skills remain independently invocable.
Resolve the epic and story paths from story_id. Check artifacts in reverse order — take the most advanced phase:
| Check | Artifact | If exists → resume at |
|---|---|---|
| 5 | stories/s{N}.{M}-retrospective.md | close |
| 4 | stories/s{N}.{M}-plan.md | implement |
| 3 | stories/s{N}.{M}-design.md | plan |
| 2 | stories/s{N}.{M}-story.md | design |
| 1 | story branch exists (story/s{N}.{M}/*) | start (branch only, no artifacts) |
| 0 | (nothing) | start (from scratch) |
Detect story type from the type field in stories/s{N}.{M}-story.md frontmatter:
| Type | Meaning | Phases skipped |
|---|---|---|
code (default) | Produces source code | None — full 8-phase chain |
docs | Documentation, guardrails, governance artifacts | AR (phase 5), QR (phase 6) |
analysis | Research, audit, assessment — no deliverable code | AR (phase 5), QR (phase 6) |
If type is missing or unrecognized, default to code.
Present: "Phase detection: resuming at {phase} (found: {artifact}), type: {type}" or "Starting fresh — no artifacts found."
Phase identified. Epic path resolved. Story type detected.Load developer profile from ~/.rai/developer.yaml. Resolve delegation level:
| Source | Resolution |
|---|---|
delegation.overrides.rai-story-run | Per-skill override (highest priority) |
delegation.default_level | Explicit default |
experience_level ShuHaRi | Shu→REVIEW, Ha→NOTIFY, Ri→AUTO |
| No profile | Default to REVIEW |
Present: "Delegation: {level}"
Delegation level resolved.Run each skill from the detected phase forward.
Phase banner (before starting):
── Phase {N}/8: {skill_name} ──
Completion banner (after finishing each phase):
Use a markdown heading + table so that file paths are clickable in the terminal:
### ✔ Phase {N}/8 — {skill_name}
| | File | Status |
|---|---|---|
| + | `path/to/new-file.md` | created |
| ~ | `path/to/modified-file.py` | modified |
**Commits:** 1 (`abc1234`) · **Tests:** 3463 passed
Rules for the completion banner:
+ for created, ~ for modified, - for deletedChain order:
| Phase | Skill | Execution | Gate after? | Skip for docs/analysis? |
|---|---|---|---|---|
| 1 | /rai-story-start {story_id} | fork | — | no |
| 2 | /rai-story-design {story_id} | fork | POST-DESIGN | no |
| 3 | /rai-story-plan {story_id} | fork | — | no |
| 4 | /rai-story-implement {story_id} | fork | POST-IMPLEMENT | no |
| 5 | /rai-architecture-review {story_id} story | fork | POST-AR | yes |
| 6 | /rai-quality-review {story_id} | fork | POST-QR | yes |
| 7 | /rai-story-review {story_id} | fork | — | no |
| 8 | /rai-story-close {story_id} | fork | — | no |
Story type skipping: When type is docs or analysis, skip phases 5 (AR) and 6 (QR) entirely. Present: "Skipping phase {N} ({skill_name}) — story type is {type}, no code to review." Then continue to the next applicable phase.
All phases fork. The orchestrator is a pure coordinator — it never executes skill logic directly. This keeps the terminal output clean (subagent output is contained) and the orchestrator context minimal.
Each fork phase runs in a fresh-context subagent via the Agent tool. This eliminates context saturation that degrades quality in later phases.
For each fork phase:
src/raise_cli/skills_base/rai-{skill-name}/SKILL.mdsubagent_type: "general-purpose"prompt: the agent prompt template below, filled with skill content and story contextAgent prompt template:
Execute the following skill for story {story_id}.
## Skill Instructions
{paste the full SKILL.md content here}
## Story Context
- Story ID: {story_id}
- Epic: {epic_id}
- Epic path: work/epics/{epic_slug}/
- Stories path: work/epics/{epic_slug}/stories/
- Prior artifacts on disk: {list each file that exists for this story, e.g. s353.2-story.md, s353.2-design.md}
## Your Task
1. Read CLAUDE.md for project-level context and rules
2. Read the prior artifacts listed above from disk
3. Execute every step in the Skill Instructions — no compression, no skipping
4. Write all output artifacts to the correct paths
5. When done, return a brief summary: what you did, artifacts created, and any verdicts or decisions
ARGUMENTS: {story_id}
Critical rules for fork execution:
Close phase (phase 8) guardrails — MANDATORY in close agent prompt:
The close agent MUST receive these explicit constraints in its prompt, in addition to the SKILL.md:
## Scope Constraints (CRITICAL — close is merge-only)
- ONLY: merge story branch, update epic scope.md, delete story branch, emit signals
- NEVER edit source code, skill files, config files, or governance docs
- NEVER create "fix" or "refactor" commits — report issues, do not repair them
- NEVER delete directories, worktrees, or files outside the story branch
- NEVER revert or modify commits already on the dev branch
- Conflict resolution: resolve ONLY the conflicting hunks mechanically — do not audit or "correct" surrounding code
- If something looks wrong, return it as a finding in your summary — do not act on it
This guardrail exists because close agents have historically rationalized unauthorized changes (removing fields, deleting directories) during conflict resolution. The close skill's "Scope Constraints" section is authoritative — these prompt guardrails reinforce it.
Each skill's SKILL.md was loaded and all its steps executed before proceeding. Skill fails → STOP immediately. Report which phase failed and why. The developer re-invokes `/rai-story-run` after fixing the issue — phase detection resumes from the last completed artifact.After phase 2 (design), phase 4 (implement), phase 5 (AR), and phase 6 (QR), apply the delegation gate:
| Level | Behavior |
|---|---|
| REVIEW | Present summary of completed phase. Wait for explicit approval before continuing. |
| NOTIFY | Present summary. Continue after 3 seconds unless user intervenes. |
| AUTO | Continue immediately. Gates still stop on test/lint/type failure or AR/QR SIMPLIFY/FAIL verdict. |
Post-design summary: Approach, components affected, key decisions. Post-implement summary: Tasks completed, tests passing, files changed. Post-AR summary: Verdict (PASS/PASS WITH QUESTIONS/SIMPLIFY), findings count, key heuristics triggered. Post-QR summary: Verdict (PASS/PASS WITH RECOMMENDATIONS/FAIL), findings count, fixes applied.
If AR verdict is SIMPLIFY or QR verdict is FAIL, STOP regardless of delegation level. Fixes must be applied before proceeding.
For docs/analysis stories, phases 5 and 6 are skipped, so their delegation gates don't apply.
After all phases complete, present:
## Story Run Complete: {story_id}
**Phases:** {start_phase} → close ({N} phases executed)
**Delegation:** {level}
**Type:** {code|docs|analysis}
**Result:** Merged to `{parent_branch}` (`{merge_commit_hash}`)
### Artifacts
| Phase | File | Op |
|-------|------|:--:|
| start | `work/epics/.../stories/s{N}.{M}-story.md` | + |
| start | `work/epics/.../stories/s{N}.{M}-scope.md` | + |
| design | `work/epics/.../stories/s{N}.{M}-design.md` | + |
| plan | `work/epics/.../stories/s{N}.{M}-plan.md` | + |
| implement | `src/path/to/file.py` | ~ |
| implement | `tests/path/to/test.py` | ~ |
| review | `work/epics/.../stories/s{N}.{M}-retrospective.md` | + |
### Metrics
| Metric | Value |
|--------|-------|
| Tests | {count} passed |
| Commits | {total_count} across {phases_count} phases |
| Patterns | {PAT-IDs or "none"} |
| Jira | {ticket} → {status} |
File paths MUST use backticks so they are clickable in the terminal. Use actual paths, not placeholders — the table above is a template.
All phases complete. Story merged and branch cleaned up.| Item | Destination |
|---|---|
| All story artifacts | work/epics/e{N}-{name}/stories/ |
| Merge commit | Parent branch (epic or dev) |
| Patterns | .raise/rai/memory/patterns.jsonl |
| Next | Next story or /rai-epic-close |
code/docs/analysis)docs/analysis stories (with message)/rai-story-start, /rai-story-design, /rai-story-plan, /rai-story-implement, /rai-architecture-review, /rai-quality-review, /rai-story-review, /rai-story-close~/.rai/developer.yaml, S325.2s325.6-design.md (decisions D1-D2-D3)