| name | phx:full |
| description | Use for large features spanning multiple contexts, new domain modules, or when the user wants autonomous end-to-end implementation. Runs the full plan-implement-review-compound cycle with specialist agents and Iron Laws enforcement. |
| effort | high |
| argument-hint | <feature description> |
Full Phoenix Feature Development
Execute complete Elixir/Phoenix feature development autonomously: research patterns,
plan with specialist agents, implement with verification, Elixir code review.
Cycles back automatically if review finds issues.
Usage
/phx:full Add user authentication with magic links
/phx:full Real-time notification system with Phoenix PubSub
/phx:full Background job processing for email campaigns --max-cycles 5
Workflow Overview
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ /phx:full {feature} โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค
โ โ
โ โโโโโโโโโโ โโโโโโโโโโ โโโโโโโโโโ โโโโโโโโโโ โโโโโโโโโโ โโโโโโโโโโ โ
โ โDiscoverโโ โ Plan โโ โ Work โโ โ Verify โโ โ Review โโ โCompoundโโDoneโ
โ โ Assess โ โ[Pn-Tm] โ โExecute โ โ Full โ โ4 Agentsโ โCapture โ โ
โ โ Decide โ โ Phases โ โ Tasks โ โ Loop โ โParallelโ โ Solve โ โ
โ โโโโโฌโโโโโ โโโโโโโโโโ โโโโโโโโโโ โโโโโฌโโโโโ โโโโโโโโโโ โโโโโโโโโโ โ
โ โ โ โ โ โ โ
โ โโโ "just do it" โโโโโโโโโโโโโค โ โ โ โ
โ โโโ "plan it" โโโ โ โ โ โ โ
โ โ โ โ โโโโโโโโโโโ โ โ
โ โ โโโโโโโโโโโโโโโโ โ โFix โโ โโโโโโโโโโโ โ โ
โ โ โ PLANNING โ โ โIssues โโโโ Fix โโโ โ
โ โ โโโโโโโโโโโโโโโโ โ โโโโโฌโโโโโ โ Review โ โ
โ โ โ โ โ Findingsโ โ
โ โ โโโโโโดโโโโโโโโโโ โโโโโโฌโโโโโ โ
โ โ โ VERIFYING โโโโโโโโโ โ
โ โโโ "research it" โโโโโโ (re-verify) โ
โ (comprehensive plan) โ
โ โ
โ On Completion: โ
โ Auto-compound: Capture solved problems โ .claude/solutions/ โ
โ Auto-suggest: /phx:document โ /phx:learn-from-fix โ
โ โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
State Machine
STATES: INITIALIZING โ DISCOVERING โ PLANNING โ WORKING โ
VERIFYING โ REVIEWING โ COMPLETED โ COMPOUNDING | BLOCKED
Save state in .claude/plans/{slug}/progress.md AND via Claude Code
tasks. Create one task per phase at start, mark in_progress on
entry and completed on exit:
TaskCreate({subject: "Discover & assess complexity", activeForm: "Discovering..."})
TaskCreate({subject: "Plan feature", activeForm: "Planning..."})
TaskCreate({subject: "Implement tasks", activeForm: "Working..."})
TaskCreate({subject: "Verify implementation", activeForm: "Verifying..."})
TaskCreate({subject: "Review with specialists", activeForm: "Reviewing..."})
TaskCreate({subject: "Capture solutions", activeForm: "Compounding..."})
Set up blockedBy dependencies between phases (sequential).
Run COMPOUNDING phase on COMPLETED to capture solved problems in .claude/solutions/.
Suggest /phx:document for docs and /phx:learn-from-fix for quick pattern capture.
Cycle Limits
| Setting | Default | Description |
|---|
--max-cycles | 10 | Max planโreview cycles |
--max-retries | 3 | Max retries per task |
--max-blockers | 5 | Max blockers before stopping |
Stop with INCOMPLETE status when limits exceeded. List remaining work and recommended action.
Integration
/phx:full = /phx:plan โ /phx:work โ /phx:verify โ /phx:review โ (fix โ /phx:verify) โ /phx:compound
Use Ralph Wiggum Loop for fully autonomous execution:
/ralph-loop:ralph-loop "/phx:full {feature}" --completion-promise "DONE" --max-iterations 50
Iron Laws
- NEVER skip verification โ Every task must pass
mix compile --warnings-as-errors before moving to the next. Run mix test <affected> per-phase, full suite only at final gate
- Respect cycle limits โ When
--max-cycles is exhausted, STOP with INCOMPLETE status. Do not continue indefinitely hoping the next fix works
- One state transition at a time โ Follow the state machine strictly. Never jump from PLANNING to REVIEWING โ each state produces artifacts the next state needs
- Discover before deciding โ Always run DISCOVERING phase to assess complexity. Skipping it for "simple" features leads to underplanned implementations
- Agent output is findings, not fixes โ Review agents report issues. Only the WORKING state makes code changes
- Skip redundant review agents โ In REVIEWING phase: skip
verification-runner (work phase already verified), skip iron-law-judge
if PostToolUse hooks verified all files. For <200 lines changed,
spawn only elixir-reviewer + security-analyzer (if auth files)
- ZERO narration in autonomous mode โ This is a HARD rule, not
a suggestion. NEVER write "Let me now...", "Now I need to...",
"I'll now...", "Next, I will...", or any preamble before a tool
call. Just call the tool. Only output text for: decisions that
need explanation, errors, or phase transitions. If you catch
yourself narrating, delete the text and just make the tool call.
(Post-PR validation: 30% of messages still violated this โ the
instruction was too soft. This stronger wording is required.)
References
${CLAUDE_SKILL_DIR}/references/execution-steps.md โ Detailed step-by-step execution
${CLAUDE_SKILL_DIR}/references/example-run.md โ Example full cycle run
${CLAUDE_SKILL_DIR}/references/safety-recovery.md โ Safety rails, resume, rollback
${CLAUDE_SKILL_DIR}/references/cycle-patterns.md โ Advanced cycling strategies