claude-code-orchestra: Skill for Claude Code

Simplify Code

Simplify and refactor $ARGUMENTS.

This skill edits source code while promising not to change behaviour and not to widen its own scope. Both promises are checked rather than asserted: Step 0 records a gate baseline and the declared target set, and Step 5 compares against it. A refactor whose baseline was never recorded cannot be reported as verified.

Declared Scope

Resolve $ARGUMENTS into a concrete list of files and/or directories before editing anything, and pass each one as a --scope entry in Step 0. That list is the answer to "was this a refactor or a rewrite" — anything else that changes is reported as out_of_scope_files and fails Step 5.

Simplification Principles

1. Single Responsibility - 1 function = 1 thing
2. Short Functions - Target under 20 lines
3. Shallow Nesting - Early return, depth ≤ 2
4. Clear Naming - Clear enough to not need comments
5. Type Hints Required - On all functions

Steps

0. Record the Baseline (mandatory, before any edit)

python3 .claude/skills/simplify/simplify_gate.py --phase before \
  --scope {target file or dir} [--scope {another}] [--base main]

It runs _shared/verify.sh and _shared/gather_diff.py and writes .claude/logs/simplify-baseline.json: the gate status of every tool as it is now, the current HEAD, the declared scope, and any file that was already modified before you started. Without this, a gate that was red the whole time is indistinguishable from a regression this refactor introduced.

Read the JSON:

• overall_before — pass / fail / no_gates. On fail, note which tools are in the baseline as failing; Step 5 will report them as pre_existing_failures rather than blaming your refactor.
• pre_existing_changes[] — dirty files you did not touch, excluded from the scope check later.
• baseline_file, artifacts[].

Exit codes: 0 baseline recorded · 1 bad arguments, or --scope missing — without a declared target set "no scope creep" cannot be checked · 2 no gate could run at all, so the refactor would be unverifiable · 3 verify.sh / gather_diff.py failed, or the baseline could not be written.