smorky850612/Aurakit / skills/aura
aura
VerifiedAll-in-one fullstack dev engine. /aura: 46 modes (build/fix/clean/deploy/review/spec/lore/ax/experiment/payment/debug/qa/orchestrate/escalate+), 6-layer security with 32 hooks, tiered models (ZERO/ECO/PRO/MAX), 8 languages, 16 specialized agents, SPEC/EARS/TRUST5/XLOOP/RALF/Autopus absorbed. ~55% token savings.
resources/autopus-pipeline-v2.md
View on GitHub# Autopus Enhanced Pipeline v2
> Absorbed from Autopus-ADK. Extends AuraKit BUILD pipeline with 5-phase structure,
> intermediate phases (1.5, 1.8, 2.1, 2.5, 3.5), and formal gate system.
---
## Full Pipeline Overview
```
Phase 1: Planning (planner agent)
Phase 1.5: Test Scaffold (tester agent — FAILING tests first)
Phase 1.8: Doc Fetch / Context7 (research enrichment)
Phase 2: Implementation (executor agents, worktree isolation)
Phase 2.1: Worktree Merge (consolidate parallel work)
Gate 2: Validation (validator — 8 checks, retry max 3)
Phase 2.5: Annotation (@AX tagging, annotator agent)
Phase 3: Testing (test runner — all must pass)
Gate 3: Coverage ≥ 85% (configurable per tier)
Phase 3.5: UX Verify (frontend changes only)
Phase 4: Review (reviewer agent, TRUST 5 criteria)
```
---
## Phase 1 — Planning
**Agent**: planner (opus in MAX/Balanced; sonnet+Amplifier in PRO; adaptive in ECO)
Steps:
1. Read SPEC (`spec.md` + `acceptance.md` if exists)
2. Assess complexity: HIGH | MEDIUM | LOW (see below)
3. Produce file manifest with agent profile assignments
4. Define implementation order (dependency sort)
5. Identify worktree candidates (parallel-safe files)
### Complexity Assessment
| Level | Criteria |
|-------|---------|
| HIGH | 3+ files changed, 200+ lines new code, OR cross-module changes |
| MEDIUM | 1-2 files, 50-200 lines, single module |
| LOW | Single file, < 50 lines, isolated change |
HIGH → opus for executor; MEDIUM/LOW → sonnet (Balanced mode only)
### Agent Assignment Table Format
```
| File | Action | Profile | Complexity | Worktree |
|------|--------|---------|------------|---------|
| src/auth/login.ts | CREATE | typescript | MEDIUM | wt-auth |
| src/auth/login.test.ts | CREATE | typescript | LOW | wt-auth |
| src/api/routes.ts | MODIFY | typescript | HIGH | wt-api |
```
---
## Phase 1.5 — Test Scaffold
**Agent**: tester (sonnet)
**Critical Rule**: Tests MUST be written BEFORE implementation.
Steps:
1. Read SPEC acceptance criteria (`acceptance.md`)
2. For each `Given/When/Then` scenario → write one test function
3. Tests MUST be failing (implementation doesn't exist yet)
4. Assert on OBSERVABLE BEHAVIOR — not just `require.NoError(t, err)`
### Behavioral Assertion Rule (Critical)
```go
// BAD — does not verify behavior
func TestCreateUser(t *testing.T) {
err := CreateUser(...)
require.NoError(t, err) // Only checks it didn't crash
}
// GOOD — asserts observable behavior
func TestCreateUser(t *testing.T) {
err := CreateUser(input)
require.NoError(t, err)
user, fetchErr := GetUser(input.Email)
require.NoError(t, fetchErr)
assert.Equal(t, input.Email, user.Email) // Verify persisted
assert.NotEmpty(t, user.ID) // ID was assigned
assert.True(t, time.Since(user.CreatedAt) < 5*time.Second)
}
```
Completion verification: All tests from Phase 1.5 MUST FAIL before Phase 2 begins.
If any test passes, tester has leaked implementation — restart.
---
## Phase 1.8 — Doc Fetch / Context7
**Agent**: haiku (read-only research)
For each library in the implementation:
1. Identify top 5 libraries used by implementation files
2. Fetch current documentation via Context7 MCP or web search
3. Store in `research.md`
**Adaptive Token Budget**:
| Libraries | Tokens per lib |
|-----------|---------------|
| 1 | 5000 |
| 2-3 | 3000 |
| 4-5 | 2000 |
| Hard cap | 10000 total |
Skip if: library is stable/well-known AND last research < 7 days old.
---
## Phase 2 — Implementation
**Agent**: executor (profile-matched)
**Rule**: ALWAYS read `spec.md` and `acceptance.md` BEFORE writing any code (Step 0).
Executor Step 0 (mandatory):
```
1. Read .autopus/specs/{SPEC-ID}/spec.md
2. Read .autopus/specs/{SPEC-ID}/acceptance.md
3. Load executor profile for file type (see autopus-executor-profiles.md)
4. Then begin implementation
```
**Worktree Isolation**: Each parallel executor works in isolated git worktree.
- Max 5 concurrent worktrees
- File ownership conflict detection before assignment
- See `autopus-worktree-isolation.md`
**Completion Report Format** (executor must return):
```
## Executor Report
Status: DONE | PARTIAL | BLOCKED
Files Modified:
- src/auth/login.ts (CREATE, 94 lines)
- src/auth/types.ts (MODIFY, +12 lines)
Notes:
- [Any deviation from plan]
- [Assumption made]
PARTIAL/BLOCKED reason:
- [What remains, what blocked]
```
**Phase 1.5 Test Constraint**: Executor MUST NOT modify test files written in Phase 1.5.
If a test needs changing, executor must flag it and stop — do not silently fix tests.
---
## Phase 2.1 — Worktree Merge
Merge all parallel executor worktrees back to main branch.
Rules:
- Merge immediately after each executor completes (sequential merge preferred)
- OR batch merge all at end of Phase 2
- Conflict → ABORT that worktree, report BLOCKED
- GC suppression: `git -c gc.auto=0 merge`
- See full rules in `autopus-worktree-isolation.md`
---
## Gate 2 — Validation
**Agent**: validator (sonnet)
**Max retries**: 3 (RALF loop — RED→GREEN→REFACTOR→LOOP)
8 validation checks (in order):
1. Build verification
2. Test execution
3. Lint / type check
4. Coverage threshold
5. File structure (300-line limit)
6. Seam verification (behavioral stub detection + smoke test)
7. Acceptance coverage (all ACs from acceptance.md mapped)
8. Migration file validation (if schema changes exist)
Gate Verdict:
```
PASS: All 8 checks passed
CONDITIONAL_PASS: Minor warnings, proceed with tracking
FAIL: One or more blocking failures — return to Phase 2
```
See full details in `autopus-validator-gates.md`.
---
## Phase 2.5 — Annotation
**Agent**: annotator (haiku)
Triggered when any of:
- File > 50 lines was created/modified
- Security-sensitive code detected
- SPEC TODOs remain incomplete
Actions:
- Add @AX:ANCHOR to public interfaces
- Add @AX:WARN to security-critical sections
- Add @AX:NOTE to non-obvious behaviors
- Add @AX:TODO for remaining SPEC items
See `autopus-ax-annotation.md`.
---
## Phase 3 — Testing
**Agent**: tester / test runner (haiku)
```bash
# Run all tests including Phase 1.5 scaffold
npm test / pytest / go test ./... / cargo test
# Coverage report
npx vitest run --coverage
pytest --cov=src --cov-report=term-missing
go test -coverprofile=coverage.out ./...
```
All Phase 1.5 tests MUST NOW PASS (implementation fills them in).
---
## Gate 3 — Coverage Threshold
| Tier | Coverage |
|------|---------|
| ECO | ≥ 70% |
| PRO | ≥ 80% |
| MAX | ≥ 90% |
| SPEC mode | ≥ 85% (fixed) |
FAIL → return to executor for additional test coverage.
---
## Phase 3.5 — UX Verify
**Agent**: ux-validator (sonnet)
**Condition**: Only runs if frontend files (*.tsx, *.vue, *.svelte) were modified.
Checks:
- Accessibility: alt attributes, label+htmlFor, role="alert" on errors
- Keyboard navigation: Tab order, Enter/Space on interactive elements
- Loading states: skeleton / spinner during async operations
- Error states: user-facing error message (not raw error object)
- Responsive behavior: mobile breakpoint coverage
Report format:
```
## UX Verification
- Accessibility: PASS / FAIL: [details]
- Keyboard Nav: PASS / FAIL
- Loading States: PASS / FAIL
- Error States: PASS / FAIL
- Responsive: PASS / SKIP (no responsive changes)
```
---
## Phase 4 — Review
**Agent**: reviewer (sonnet in ECO/PRO; opus in MAX)
TRUST 5 review criteria:
- **T**ested: Are tests comprehensive? Behavioral assertions?
- **R**eadable: Can another agent/dev understand this in 30 seconds?
- **U**nified: Consistent with project conventions and patterns?
- **S**ecured: No injection, no hardcoded secrets, proper auth?
- **T**rackable: Lore trailers present? Decision rationale captured?
Parallel review pattern (MAX tier):
```
Reviewer-A (code quality) ──┐
Reviewer-B (security) ──┤→ Lead Consolidator → unified report
Reviewer-C (architecture) ──┘
```
REQUEST_CHANGES triggers retry (max 2 retries in Phase 4).
---
## RALF Loop (Quality Gate Retry)
```
RED → Gate failure detected
GREEN → Builder fixes the specific issue
REFACTOR → Cleanup if fix introduced new issues
LOOP → Re-run validation
```
Abort conditions:
- Gate 2: 3 consecutive failures → escalate to user
- Phase 4: 2 consecutive REQUEST_CHANGES → escalate
- Circuit breaker: 3 failures with same error → STOP
---
## AuraKit Integration
Use `/aura build: [feature] --pipeline=v2` to activate full pipeline.
Or configure as default in `.aura/config.json`:
```json
{
"pipeline": "v2",
"coverage_threshold": 85,
"lore_commits": true,
"ax_annotation": true
}
```
Default in SPEC mode: always uses pipeline v2.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |