ShaheerKhawaja/full-cycle

full-cycle

The nuclear option. Chains 7 skills into a complete audit-research-plan-fix-verify-ship pipeline. This is the most comprehensive ProductionOS composite skill -- it touches every layer of the codebase and produces a fully evaluated, reviewed, and shipped result.

Cost: High. Expect 30-90 minutes of agent time depending on codebase size. When to use: Major milestones, pre-launch readiness, end-of-sprint quality gates, or when the answer to "what needs work?" is "everything." When NOT to use: Quick fixes, single-file changes, debugging a specific bug. Use targeted skills instead.

Chain Overview

security-audit -> production-upgrade -> deep-research -> plan-ceo-review
      |                  |                   |                 |
      v                  v                   v                 v
  AUDIT.md           UPGRADE.md         RESEARCH.md       CEO-REVIEW.md
      |                  |                   |                 |
      +------ Phase 1: Assess -----------+  +-- Phase 2: Plan -+
                                                               |
                                                               v
auto-swarm ---------> self-eval ---------> ship
      |                  |                   |
      v                  v                   v
  SWARM-RESULT.md    EVAL.md             PR created
      |                  |
      +-- Phase 3: Fix --+--- Phase 4: Ship ---+

Inputs

| Parameter | Values | Default | Description | |-----------|--------|---------|-------------| | target | path, scope | . | What to process | | depth | medium, deep, ultra | deep | Swarm depth for fix phase |

---

Phase 1: Assess (Steps 1-2)

Step 1: Security Audit

Invokes: /security-audit with framework=all, scope=full

Runs the full 7-domain security scan. Produces .productionos/AUDIT-SECURITY.md.

Escalation: If CRITICAL findings > 0, the full-cycle continues but flags them as MUST-FIX items for the swarm phase. Unlike audit-and-fix, full-cycle does not halt on CRITICALs -- it feeds them into the fix pipeline.

Step 2: Production Upgrade

Invokes: /production-upgrade

Full codebase quality audit beyond security: code style, error handling, test coverage, dependency health, API design, performance, maintainability. Produces .productionos/UPGRADE.md.

Phase 1 output: Combined issue list from security audit + production upgrade, deduplicated and priority-ranked.

---

Phase 2: Plan (Steps 3-4)

Step 3: Deep Research

Invokes: /deep-research

Researches solutions for the top issues found in Phase 1. Cross-references ~/repos/ for existing implementations, checks library docs via context7, identifies best-practice patterns. Produces .productionos/RESEARCH.md.

Step 4: CEO Review

Invokes: /plan-ceo-review

Applies founder lens to the research findings. Identifies the 10-star fix strategy, scope expansions worth taking, and priority ordering. Produces .productionos/CEO-REVIEW.md.

Phase 2 output: Prioritized fix plan with CEO-approved scope, ready for automated execution.

---

Phase 3: Fix (Step 5)

Step 5: Auto-Swarm

Invokes: /auto-swarm --mode=fix

Parallel agent swarm that executes the fix plan from Phase 2. Each agent targets one issue or improvement. Swarm operates with quality gates -- each fix must pass its own self-check before merging.

Swarm configuration:

• Depth: {depth} parameter (default: deep)
• Mode: fix (apply changes, not just audit)
• Quality gate: each fix self-evaluated, >= 8.0 required
• Rollback: any fix that causes test regression is reverted

Produces: .productionos/SWARM-RESULT.md

Estimated effort by depth: | Depth | Agents | Time | Coverage | |-------|--------|------|----------| | medium | 5-15 | 15-30 min | Top priority issues | | deep | 15-50 | 30-60 min | All HIGH+ issues | | ultra | 50-100+ | 60-90 min | All issues including LOW |

---

Phase 4: Ship (Steps 6-7)

Step 6: Self-Evaluation

Invokes: /self-eval session

Evaluates the entire session: all changes made across all phases. Scores quality, necessity, correctness, completeness. This is the final quality gate before shipping.

Quality gate: Score >= 8.0 required. If < 8.0 after 3 self-heal loops, BLOCK shipping and report remaining issues.

Step 7: Ship

Invokes: /ship

Ships the result: syncs with base branch, runs tests, bumps version, updates changelog, commits, pushes, creates PR.

Pre-ship gate: Only runs if Step 6 scored >= 8.0 AND tests pass.

---

Abort Conditions

The full-cycle aborts (does not ship) when:

| Condition | Aborts At | Recovery | |-----------|-----------|----------| | Self-eval < 6.0 after 3 loops | Step 6 | Fix manually, re-run | | Tests fail after swarm fixes | Step 5 | Rollback, investigate | | Merge conflicts with base | Step 7 | Resolve conflicts, re-run Step 7 | | > 3 CRITICAL unfixed after swarm | Step 5 | Manual CRITICAL fix required |

---

Output Format

Final composite report written to .productionos/FULL-CYCLE.md:

# Full Cycle Report

## Summary
- **Security Posture:** X/10 (before) -> Y/10 (after)
- **Code Quality:** X/10 (before) -> Y/10 (after)
- **Issues Found:** N total
- **Issues Fixed:** M fixed, K remaining
- **Self-Eval Score:** X/10
- **Ship Status:** SHIPPED (PR #{N}) | BLOCKED

## Phase Results
| Phase | Step | Status | Output |
|-------|------|--------|--------|
| Assess | Security Audit | DONE | N findings |
| Assess | Production Upgrade | DONE | M opportunities |
| Plan | Deep Research | DONE | K solutions found |
| Plan | CEO Review | DONE | Scope approved |
| Fix | Auto-Swarm | DONE | J fixes applied |
| Ship | Self-Eval | PASS/FAIL | X/10 |
| Ship | Ship | DONE/BLOCKED | PR #{N} |

## Before/After Comparison
| Metric | Before | After | Delta |
|--------|--------|-------|-------|
| Security score | X | Y | +Z |
| Test coverage | X% | Y% | +Z% |
| Open issues | N | M | -K |

## Remaining Items
{issues not addressed in this cycle}

Full cycle completed: {timestamp} | Duration: {minutes}m

---

Cost Estimates

| Codebase Size | Estimated Duration | Agent-Minutes | Recommended Depth | |---------------|-------------------|---------------|-------------------| | Small (< 10K LOC) | 15-25 min | ~50 | deep | | Medium (10-50K LOC) | 30-50 min | ~150 | deep | | Large (50-200K LOC) | 45-75 min | ~300 | medium | | Huge (> 200K LOC) | 60-90 min | ~500 | medium |

When to Use

• "Make this production-ready" -- full 7-step pipeline
• "Do everything" -- full 7-step pipeline
• "End-to-end quality pass" -- full 7-step pipeline
• "Pre-launch checklist" -- full 7-step pipeline

When NOT to Use

• Fixing one bug -- use /debug
• Shipping one feature -- use /ship-safe
• Just auditing -- use /audit-and-fix
• Just planning -- use /research-and-plan
• Quick marketing check -- use /growth-audit