copilot-instructions.md · git:20261011.d65ea2c · 2026-10-11 · sha256 1d11ffdcd655c30b
copilot-instructions.md git:20261011.d65ea2cA
Immutable. This exact content never changes and is served at /api/v1/blob/1d11ffdcd655c30b.
# nobrainer-tech-flow repository instructions for Copilot The eighteen canonical Agent Skills are under `skills/`. Read the relevant `skills/<name>/SKILL.md` before changing a skill; retired predecessors exist only in Git history. For non-trivial delivery load `nobrainer-tech-flow`. Inspect current state, clarify once only when the answer changes scope or safety, freeze the minimum change and show a compact Progress checklist. Keep one-step work direct. Use a detailed ledger only for multi-session, dependency-rich, consequential or explicitly resumable work. Delegate every independent ready unit to native subagents within verified capacity and quota; MAIN coordinates, integrates and verifies. Team supplies capabilities; Dispatcher schedules real queues. Sessions operates authorized visible sessions. Audit reports before state advances and refill freed slots. `PROBLEM_GATE`: begin with the literal failure, local evidence and smallest reproducer. Query related wiki decisions when useful. Use current internet research only when the remedy depends on an external, current, niche, uncertain or high-stakes fact. If required research is unavailable, return `RESEARCH_BLOCKED` and choose no remedy that depends on it. Keep `SKILL.md` portable: line-1 YAML frontmatter with only `name` and `description`, directory/name match, relative links, no secrets, private paths, private client data, hardcoded drifting model versions or unsafe shell placeholders. Run `python3 scripts/validate_skills.py --suite` and relevant tests before reporting success. Use `nobrainer-skill-doctor` for portfolio-wide instruction, skill or task-prompt audits. Use `nobrainer-build` for implementation, `nobrainer-security` for material trust boundaries and `nobrainer-review` for a justified final evidence gate.