Develop

# Develop

@description Scaffold, implement, test, and iterate on a strategy until it passes QA.
@arguments $DESIGN_SPEC: Path to the strategy design specification (from /discover or provided directly)

Load agents: dev, qa.
Load skills: canon-conventions, backtesting, orchestrator, risk-management.

Run every step below in order. Do not stop or ask for confirmation between steps.

**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. Verify scaffold

As dev, verify the project scaffold exists before implementing.

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=scaffold status=running log.info="Verifying scaffold..."

Check that these files are present:

• package.json
• tsconfig.json
• src/types/TradeSignal.ts
• src/types/RiskInterface.ts
• .canon/dega-core.yaml
• .canon/config.yaml
• AGENTS.md

If any are missing, stop and tell the user:

Project scaffold is incomplete. Run /canon-init first to set up the Canon framework, then re-run /develop.

2. Implement

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=develop status=running log.info="Implementing strategy..."

Implement strategy logic from the design specification.

Load skills: canon-conventions, risk-management.

Required:

• Implement TradeSignal interface in src/strategy.ts
• Implement RiskInterface in src/types/RiskInterface.ts with the hard limits from the design specification
• Follow domain layering: Types → Config → Repo → Service → Runtime → UI
• Use agent-oriented error messages (what/why/how format)

Do not skip RiskInterface. "I'll add it later" is not acceptable.

3. Test

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=test status=running log.info="Running checks..."

Run every check command from .canon/dega-core.yaml success_criteria:

pnpm exec tsc --noEmit
pnpm exec oxlint src/
pnpm exec vitest run

All three must pass. If .canon/dega-core.yaml has additional strategy-specific checks beyond these defaults, run those too.

Review test results against the design spec's success criteria — tests passing is the minimum bar, also check backtest metrics if applicable.

4. Iterate (Ralph Loop)

If any checks from step 3 fail, or backtest criteria from the design spec are not met, iterate:

1. Read the failing check output to identify what broke.
2. Fix the issue in code.
3. Re-run all check commands from .canon/dega-core.yaml success_criteria.
4. Repeat until all checks pass.

Load orchestrator skill for iteration guidance.

The max_iterations in .canon/dega-core.yaml limits how many cycles to attempt. If the limit is reached without meeting all criteria, surface the specific failing criteria for human review before proceeding.

5. QA review

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=qa status=running log.info="QA review..."

As qa, validate strategy quality.

Load skills: canon-conventions, backtesting, risk-management.

Check:

1. Code conventions: Domain layering respected, error messages follow what/why/how
2. Backtest results across multiple timeframes (7d, 30d, 90d if data available)
3. No overfitting signals (parameter stability, out-of-sample test)
4. RiskInterface correctly enforces hard limits (not just present — verify logic)
5. Edge cases: zero liquidity, API timeout, zero balance

Verdict:

• Approved: All criteria met → output QA-approved summary, proceed to register workflow
• Return to dev: Specific blocking issues found → list each issue with severity (blocking / advisory) and suggested fix → loop back to step 2

Completion criteria

• Tests pass (pnpm exec vitest run)
• Lint clean (pnpm exec oxlint src/)
• Types valid (pnpm exec tsc --noEmit)
• Backtest criteria from design spec met
• QA approved (≥30 trades, profit_factor ≥1.0, no blocking biases)
• QA-approved strategy ready to hand off to register workflow