brewtoolsskill-togglelisted

# Skill Toggle > **Disable/enable individual plugin skills** via the stable `skillOverrides` mechanism in `~/.claude/settings.json` (Claude Code 2.1.115+). State survives plugin updates — **no SessionStart reapply hook needed**. See also: `/brewtools:agent-toggle` for agents (uses native `permissions.deny`). <instructions> ## Robustness Rules | Rule | Applies | |------|---------| | Every Bash call ends with `&& echo "OK ..." \|\| echo "FAILED ..."` | ALL | | Never use `Write`/`Edit` on `~/.claude/settings.json` — use Bash + Node helper | ALL | | This skill mutates **skills only** (`skillOverrides`) — never agents | ALL | | Atomic writes via lockfile + temp + rename (handled by helper) | P2 | Paths (substitute literally in Bash): - Helper: `$CLAUDE_PLUGIN_ROOT/skills/skill-toggle/helpers/overrides.mjs` - Settings: `~/.claude/settings.json` (resolved to `$HOME/.claude/settings.json`) - Legacy global state (read-only, backwards visibility): `$CLAUDE_PLUGIN_DATA/toggle-state.json` ## Override Modes | Mode | Effect | |------|--------| | `off` | Fully disabled — invisible to user and LLM | | `user-invocable-only` | Only via `/plugin:skill` slash; LLM cannot auto-invoke | | `name-only` | Name visible in autocomplete, instructions/body not loaded | | `on` | (Re-enable) — entry removed from settings.json | Default for "disable" intent without explicit mode: `off`. --- ## Phase I — Interactive Flow (entry gate) > **Full spec:** `_shared/toggle/interactive-flow.md` (phases I0