SDLC Enforcement
August 16, 2026 ยท View on GitHub
Before Every Task
- Plan before coding - outline steps, state confidence (HIGH/MEDIUM/LOW)
- LOW confidence? Research more or ASK USER
- Reasoning policy - use
gpt-5.6-solwithhighreasoning for this repo - Always keep this repo on
gpt-5.6-solhigh; do not switch wizard-repo work tomixed, Terra, Luna, or lower-tier profiles - Keep this repo on
maximum(gpt-5.6-solhigh) because codex-sdlc-wizard is unusually meta and high-blast-radius - Default meaningful agentic coding to Sol
high; escalate only difficult or high-risk slices toxhighwhenhighleaves unresolved risk - Max is a single-task reasoning escalation; Ultra is a subagent-backed parallel-work escalation. Most tasks do not need Max or Ultra, and neither is a default for this repo.
- If
GOALS.mdexists, treat it as the active-scope contract and keepROADMAP.mdas backlog/history - Use RED only when a RED mutation is writable; otherwise choose an honest eval, mechanical assertion, or review gate
- ALL tests must pass before commit - no exceptions
Evidence Workflow (MANDATORY)
- EVAL it when agent-facing guidance has an observable scenario.
- Plain-assert it for a mechanical contract.
- Review it for judgment-call prose; cross-model review is the guard.
- For executable behavior, any observable input/output or side-effect difference makes a RED mutation writable; the meaning exception applies only to prose judged by a reader.
- Implement-first is allowed only when a named gate blocks the required RED or evidence act itself. A gate refusing implementation because RED is missing is working, not an entry ticket. A cross-model ruling must APPROVE that same act and scope before the edit.
- Only commit after the selected evidence is green.
After Implementation
- Self-review: read back your changes, check for bugs
- Run full test suite - ALL tests must pass
- Only then: commit and push
AI Setup Lanes
This repo ships and uses Sol high as the normal root driver, with task-scoped xhigh escalation. The Terra-led mixed profile is experimental explicit opt-in, and Terra/Luna otherwise stay bounded support options. The root agent normally owns planning; specialist agents and Ultra are optional when the task benefits. See AI_SETUP_LANES.md.
Rules
- Delete legacy code - no backwards compatibility hacks
- Less is more - don't add what wasn't asked for
- Tests ARE code - treat test failures as bugs
- NEVER commit without running tests first
- During setup, environment repair, and auth-heavy workflows, prefer full access