teddytennant

Write Architecture Decision Records when choosing between real alternatives.

When the agent loop is stuck: capture, shrink, recover.

REST/JSON API consistency: resources, errors, pagination.

Structured review output with severities and explicit approval gate.

Systematic debug: reproduce, isolate, fix, prevent.

Consistent error strategy across layers.

Branch, commit, and PR conventions without ceremony.

Measure before optimizing; avoid cargo-cult perf.

Baseline coding workflow. Explore before editing, make minimal precise diffs, verify with the project's own build and tests, clean deliverable directories, and review the diff before finishing.

How to extend Wizard via /evolve. Cheapest tier first — skills, MCP, scripted tools, subagents; deep recompile only when the binary must change.

Wizard's evolved coding workflow. Use when writing or modifying code, fixing bugs, or delivering to an explicit spec (required paths, report schemas, exit codes) — and for its battle-tested recipes on long-lived services, native/Cython package installs, polyglot builds, and image/board/puzzle analysis. Explore before editing, minimal precise diffs, verify with the project's own tests, clean deliverable trees, review the diff before finishing.

How to extend Claude Code's own capabilities, Wizard-style. Use when asked to add a new capability, tool, skill, MCP server, or subagent to Claude Code itself — or when a task needs a capability the harness does not yet have (browser, database, computer use, a domain workflow). Pick the cheapest channel that works and verify it is live before reporting success.

Baseline coding workflow. Explore before editing, make minimal precise diffs, verify with the project's own build and tests, clean deliverable directories, and review the diff before finishing.

How to extend Wizard's own capabilities via /evolve. Picking the cheapest tier, the four runtime channels (skills, MCP servers, scripted tools, subagents), and when a deep recompile is justified.

Thorough whole-codebase audit using subagents. Fans out reviewer agents across every subsystem, has independent skeptics adversarially verify each finding, and writes a ranked report to AUDIT.md. Use when the user wants to audit, review, or find bugs/security issues across an entire existing codebase (not just a diff) — e.g. "audit this repo", "review the whole codebase", "find bugs everywhere", "security audit". For reviewing a branch/PR diff, prefer /code-review instead.

Use when a coding task spans many steps, files, or a long session: imposes a phased execution loop, evidence-driven verification, and tracked falsifiable progress so long-horizon work doesn't drift, stall, or silently regress.