Ralph Cycle

@description Autonomous iteration loop — execute, check, iterate until all success criteria pass or budget is exhausted.

maulana-tech updated 1mo ago
Claude CodeGeneric
View source ↗
# Ralph Cycle

@description Autonomous iteration loop — execute, check, iterate until all success criteria pass or budget is exhausted.

Load agent: dev.
Load skills: orchestrator, canon-conventions.

Run every step below. The loop repeats steps 1-3 until the SHIP or ESCALATE condition is met.

**State write convention:** Every `terminal-ui-write.sh` call below is guarded — check
if the script exists and skip silently if not. Do not warn about missing dashboard.

## 1. Execute

Write state update:

```bash
TUI_WRITE="${DEGA_CORE_HOME:-${HOME}/.degacore}/scripts/terminal-ui-write.sh"
[[ -f "${TUI_WRITE}" ]] && \
  bash "${TUI_WRITE}" .canon/state.json \
    phase=execute status=running log.info="Iteration ${ITERATION:-1}: executing..."

Implement or modify code toward the success criteria defined in .canon/dega-core.yaml.

Load skills: canon-conventions, orchestrator.

Input:

  • Task description
  • Success criteria from .canon/dega-core.yaml
  • Feedback from previous iteration (if any)

Apply changes. Follow domain layering and error message conventions.

2. Check

Write state update:

TUI_WRITE="${DEGA_CORE_HOME:-${HOME}/.degacore}/scripts/terminal-ui-write.sh"
[[ -f "${TUI_WRITE}" ]] && \
  bash "${TUI_WRITE}" .canon/state.json \
    phase=check status=running log.info="Running success criteria checks..."

Run every check command listed in .canon/dega-core.yaml under success_criteria.

For each entry, run the check command (e.g., pnpm exec tsc --noEmit, pnpm exec oxlint src/, pnpm exec vitest run).

Record each criterion: pass ✓ or fail ✗ with the specific failure message.

3. Analyze (if checks failed)

If any check failed, analyze the failures before the next iteration:

  1. Identify root cause of each failure
  2. Plan the fix — be specific (which file, which line, what change)
  3. Estimate confidence: Is this fix likely to resolve the failure?

If stuck (same failure across 3+ iterations without progress), escalate immediately rather than burning more budget.

4. Loop or ship

Evaluate the current state:

SHIP if:

  • All success criteria pass → Write state update and summary, exit loop:
TUI_WRITE="${DEGA_CORE_HOME:-${HOME}/.degacore}/scripts/terminal-ui-write.sh"
[[ -f "${TUI_WRITE}" ]] && \
  bash "${TUI_WRITE}" .canon/state.json \
    status=idle log.info="SHIP — all criteria met"

LOOP if:

  • Any criterion failed AND iteration count < max_iterations AND budget remaining → Apply the fix plan from step 3, return to step 1

ESCALATE if:

  • Iteration count ≥ max_iterations, OR
  • Budget exhausted (max_tokens or max_spend reached), OR
  • Stuck (same failure ≥3 iterations without progress)
TUI_WRITE="${DEGA_CORE_HOME:-${HOME}/.degacore}/scripts/terminal-ui-write.sh"
[[ -f "${TUI_WRITE}" ]] && \
  bash "${TUI_WRITE}" .canon/state.json \
    status=error error="Escalated — human intervention required"

→ Document: which criteria are failing, what was tried, why it's stuck → Surface to human for intervention

Completion criteria

  • SHIP: All success criteria pass
  • ESCALATE: Budget/iteration limit reached, or stuck — with full failure report for human review

Maintain Ralph Cycle?

Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.

[Ralph Cycle on getagentictools](https://getagentictools.com/loops/maulana-tech-ralph-cycle?ref=badge)