schlenks/rule-of-five-code

Rule of Five — Code

Each pass has ONE job. Re-read the entire artifact through that lens. See references/pass-order-rationale.md for order rationale and convergence details.

Quick Start

Create native tasks for 5 passes with sequential dependencies:

TaskCreate: "Pass 1: Draft"
  description: "Shape and structure. Get the outline right. Breadth over depth."
  activeForm: "Drafting"

TaskCreate: "Pass 2: Correctness"
  description: "Logic, bugs, regressions. Does it work? Did it break anything?"
  activeForm: "Checking correctness"
  addBlockedBy: [draft-task-id]

TaskCreate: "Pass 3: Clarity"
  description: "Comprehension. Can someone unfamiliar understand this? Simplify."
  activeForm: "Improving clarity"
  addBlockedBy: [correctness-task-id]

TaskCreate: "Pass 4: Edge Cases"
  description: "Failure modes. What's missing? What breaks under stress?"
  activeForm: "Handling edge cases"
  addBlockedBy: [clarity-task-id]

TaskCreate: "Pass 5: Excellence"
  description: "Pride. Would you show this to a senior colleague? Polish rough spots."
  activeForm: "Polishing"
  addBlockedBy: [edge-cases-task-id]

ENFORCEMENT:

• Each pass is blocked until the previous completes
• Cannot commit until all 5 tasks show status: completed
• TaskList shows your progress through the passes
• Skipping passes is visible - blocked tasks can't be marked in_progress

Cross-Model Review (Codex)

Check availability: Look for <codex-integration> in the session context (injected by session-start hook). If absent, skip this section entirely.

If present, extract the install path from the tag. Use this as {RESOLVED_CODEX_PATH} — embed it literally in the agent prompt below. No Bash commands needed. Then dispatch:

Agent:
  run_in_background: true
  description: "Codex cross-model audit (code)"
  prompt: |
    Run a Codex adversarial review of the current changes.

    ```bash
    node "{RESOLVED_CODEX_PATH}/scripts/codex-companion.mjs" adversarial-review --wait
    ```

    Persist the full output to a timestamped temp file (background agent messages may be truncated):
    ```bash
    mkdir -p temp
    AUDIT_TS=$(date +%Y%m%d-%H%M%S)
    tee temp/codex-audit-code-${AUDIT_TS}.md <<'CODEX_AUDIT_EOF'
    [full codex review output]
    CODEX_AUDIT_EOF
    ```

    Output the full review as your final message.

This runs concurrently with all 5 passes — zero blocking. Codex uses auto-detect scope: reviews uncommitted changes if working tree is dirty, or branch diff against default branch if clean (e.g., after SDD implementer commits).

After pass 5 completes, wait for the Codex background agent to finish before presenting results. Do NOT present pass 5 results until the Codex review has either completed or timed out. This is a synchronous gate — the rule-of-five skill does not have a monitor loop or late-delivery mechanism, so all output must be collected before the skill finishes.

• If Codex completed successfully: Read the Codex agent's persisted temp file (primary) or fall back to agent output. Present as "Cross-Model Audit (Codex)" section after pass 5 results.
• If Codex failed or timed out: append _Codex cross-model audit was unavailable for this run._ after pass 5 results

## Cross-Model Audit (Codex)

[Full Codex adversarial review output — verdict, findings, recommendations]

For each pass: re-read the full artifact, evaluate through that lens only, make changes, then mark task complete.

Detection Triggers

Invoke when: >50 lines of code written/modified, new public API/interface/component, complex multi-file refactoring, security-sensitive changes, or about to claim code work "done".

For plans/design docs/skills, use rule-of-five-plans. For tests, use rule-of-five-tests.

Skip for: Single-line fixes, typo corrections, trivial changes under 20 lines.

Announce: "Applying rule-of-five-code to [artifact]. Starting 5-pass review."

The 5 Passes

| Pass | Focus | Exit when... | |------|-------|--------------| | Draft | Shape and structure. Don't perfect - get the outline right. Breadth over depth. | All major components exist | | Correctness | Logic, bugs, regressions. Does it work? Did it break anything? | No known errors; no regressions | | Clarity | Comprehension. Can someone unfamiliar understand this? Simplify. Cut jargon. | A newcomer could follow it | | Edge Cases | Failure modes. What's missing? What breaks under stress? | Unusual inputs handled | | Excellence | Pride. Would you show this to a senior colleague? Polish the rough spots. | You'd sign your name to it |

Common Mistakes

| Mistake | Fix | |---------|-----| | Multiple lenses in one pass | ONE lens per pass. Correctness pass ignores style. | | Skipping passes on "simple" artifacts | All 5 or none. Choose a different approach for trivial work. | | Rushing through passes | Each pass: genuinely re-read the full artifact | | Pass finds nothing to change | That's fine. Move on. Not every pass surfaces issues. | | Applying to entire codebase | Artifact = the unit you're changing. A function, component, or document--not the whole system. | | Ignoring consumers when modifying | Grep for usages. Check if callers depend on specific behavior you're changing. |

Modifying Existing Code

Shift Correctness pass: "Did I break anything?" matters more than "Does my addition work?"

Correctness checklist for modifications:

1. Does my change work correctly?
2. Did I break the code I modified?
3. Did I break tests that depend on old behavior?
4. Did I break consumers? (Other code that calls/uses what I changed)

See references/modification-checklist.md for interface changes and consumer breakage signs.

Reference Files

• references/example-before-after.md: Full before/after code example with findings table
• references/pass-order-rationale.md: Order rationale, convergence theory, cross-pass issues
• references/modification-checklist.md: Interface change checklist and consumer breakage signs

<!-- compressed: 2026-02-11, original: 700 words, compressed: 547 words -->