init-rules

Init Rules

Overview

Install curated development rules based on detected project context. Two delivery modes:

• --target=rules (default) — write rule files into .claude/rules/ (project) or ~/.claude/rules/ (global). Files with paths: frontmatter stay path-scoped.
• --target=claudemd — inject rule content directly into ./CLAUDE.md (project) or ~/.claude/CLAUDE.md (global), wrapped in sentinel blocks for idempotent re-runs and clean uninstall.

When to Use

Triggers: "init rules", "set up rules", "install rules", "configure coding rules", "onboard project".

When NOT to use

• The user wants to edit rule content itself — open the rule file directly.
• The user wants per-file rule scoping — use --target=rules; claudemd mode cannot path-scope.
• The target directory or CLAUDE.md is managed by another tool and should not be touched.

Arguments

• project (default) → scope to working directory
• global → scope to ~/.claude/
• --target=rules (default) → write as individual rule files
• --target=claudemd → inject into CLAUDE.md with sentinel blocks
• --force → overwrite existing rule files / replace existing sentinel blocks
• --uninstall → remove rules previously installed by this skill (mode-aware)

Parse from $ARGUMENTS. Unknown or conflicting tokens → AskUserQuestion; never guess.

Core Process

Phase	Goal	Details
0	Setup + user approval	Inline below
1	Install / update / uninstall	phases/run.md

Phase 0: Setup

1. Parse $ARGUMENTS: scope (project/global), target (rules/claudemd), --force, --uninstall. Unknown token → AskUserQuestion.
2. task_name = <YYYY_MM_DD>_init_rules_<scope>_<target> where <YYYY_MM_DD> is today's date (underscores); on same-day collision append _v2, _v3.
3. Create .mz/task/<task_name>/.
4. Write state.md with Status: running, Phase: 0, Started: <ISO>, Scope: <project|global>, Target: <rules|claudemd>, Force: <bool>, Uninstall: <bool>, DetectedContexts: [], Installed: [], Replaced: [], Skipped: [], Removed: [].
5. Emit setup block: task_name, resolved target path, mode flags.

For --target=claudemd installs, the approval gate in Phase 1 Step 4b must run before any write. A single run-level confirmation covers all subsequent writes in the run.

Read phases/run.md and proceed to Phase 1.

Techniques

Techniques: delegated to phase files — see Phase Overview table above.

Common Rationalizations

N/A — collaboration/reference skill.

Red Flags

Red Flags: delegated to phase files — see Phase Overview table above.

Verification

Verification: delegated to phase files — see Phase Overview table above.

Error Handling

• Empty / ambiguous argument → AskUserQuestion; never guess.
• Missing tooling — if ${CLAUDE_PLUGIN_ROOT} is unset or the bundled rules/ directory is not readable, AskUserQuestion with the failing path.
• No detected contexts → still install universal rules; if target not writable, retry once then AskUserQuestion.
• CLAUDE.md unreadable or unwritable → AskUserQuestion with the exact path and error; do not fall back silently.
• Duplicate sentinel IDs found in CLAUDE.md (manual edit or prior bug) → stop, report the conflicting lines, ask the user to resolve before re-running.
• Plugin version unreadable → use literal unknown for the version tag and note it in the report.
• Never guess — any ambiguity escalates via AskUserQuestion rather than silent overwrite or skip.

State Management

After each phase, update .mz/task/<task_name>/state.md with:

• Status: running | complete | aborted_by_user | failed
• Phase: current phase number
• DetectedContexts: list of detected language/tooling tags
• Installed: / Replaced: / Skipped: / Removed: filenames or rule ids touched in this run

Never rely on conversation memory for cross-phase state — context compaction destroys specific paths and decisions.