Audit and surgically fix the marketplace's documentation so every shipped command is discoverable — checks each plugin README's Skills section against SKILL.md frontmatter, the root README's plugin list and command index, manifest/description sync, and docs-site parity. Use when docs may have drifted from the code…
Apply progressive-disclosure discipline to a SKILL.md — measure size, identify inert content (prompt templates, schemas, long enumerations, long bash blocks), propose extractions, and rewrite the SKILL.md with reference stubs. Use when editing or creating a SKILL.md, or when an existing SKILL.md exceeds 100 lines.
Set up, reset, or run a local testbed for the plugins in this repo. Use whenever you want to dogfood a plugin from this working copy (any branch) against a clean sibling sandbox project — covers marketplace registration, install/update, per-plugin smoke recipes.
Read-only inspector for saved handoffs. Reports whether .claude/handoff/active.md exists, its age, saved-at SHA/branch, progress past the save point (commits/branch changes), snooze status, and current settings. Use when the user asks about handoff state, why a handoff isn't surfacing, or wants a quick sanity check…
Use when the user explicitly wants to discard a saved handoff without acting on it (the work is no longer relevant). Invokable anywhere via /handoff:discard.
Use when a saved handoff is surfaced at SessionStart, or when the user runs /handoff:pickup, to resume prior work — presents the saved directive and dispatches on the user's choice.
Use when the user signals end-of-session — "save handoff", "pick this up later", "continue in a new session", "I'll come back to this". Captures the session's intent so the next session can resume it. Slash form /handoff:save is the explicit fallback.
Internal capability-declaration surface for a programmatic caller's compatibility gate — answers "which lens is installed, and does it support X?" before any review is sent. Returns the manifest version, the advertised capability tokens, and the input keys and resolution rules a caller may rely on; a requested token…
Internal data-only review core invoked directly by /lens:review and by a programmatic orchestrator (matali) calling it as a compute-only step, to produce a review-findings JSON object from the session diff. Not user-invocable; returns data, writes nothing, never prompts.
Internal pure-render entrypoint — given a review-findings object (plus optional prior findings, a diff ref, and intent), reconcile in-memory, assemble the review-model, and render the interactive HTML via walkthrough:render. Writes ONLY the output HTML; no lens state, no recompute, no task list. Invoked by an…
Review the current session's changes against their spec and plan, adversarially verify the findings, and render an interactive review document. Use when the user wants to review what was just built before it ships, runs /lens:review, or asks to "review my changes / review this before I commit". Reviews Claude's own…
Use only when the user explicitly invokes /notify:setup — one-time setup of cross-platform desktop notifications (installs the notifier, wires the Stop/Notification/SubagentStop hooks). User-invoked only.
Use only when the user explicitly invokes /notify:uninstall — removes notify's hooks and config while leaving other settings intact. User-invoked only.
Interactive Q&A for customizing notification preferences (sounds, durations, matcher regex, per-event toggles). Internal building block invoked by the notify setup skill — not user-invocable.
Bring a repo's existing hand-crafted Claude tooling under onboard management — synthesizes an onboard baseline (retrofit-mode meta + snapshots) WITHOUT modifying any hand-crafted file, so /onboard:update works afterward. Use only when the user explicitly invokes /onboard:adopt, or when start/update route here.
Apply accumulated tooling drift to Claude configuration. Use when user says their Claude tooling is out of sync, mentions drift detected by hooks, installed/removed a plugin and wants the Plugin Integration section refreshed, or asks to drain queued drift-log updates.
Internal v3 generation step — turns the onboard:start v3 context (analysis + wizardAnswers + research) into Claude tooling artifacts via the config-generator agent. Invoked through the Skill tool by /onboard:start and /onboard:update; not an external API; not user-invocable.
Invoked by the config-generator agent during /onboard:start and /onboard:generate to produce the Claude tooling artifacts. Internal building block; not user-invocable.
Internal research engine for onboard v3 — fans out read-only specialists per dimension, adversarially verifies their claims, synthesizes the canonical research object, and renders four markdown artifacts. Invoked via the Skill tool by /onboard:start or directly by a model; not user-invocable. Returns the validated…
Use ONLY when the user explicitly runs /onboard:start. The full interactive onboarding wizard that sets up tailored Claude Code tooling for a project. Never auto-invoke.
Update existing Claude tooling against latest best practices — re-analyzes the codebase, fetches current Claude Code documentation, presents targeted upgrade options, and applies approved changes with merge/replace choice per file. Use only when user explicitly invokes /onboard:update.
Independent feature verification for projects with feature tracking. Use when user wants to test whether features in docs/feature-list.json pass their acceptance criteria, check sprint gate status, or run verification on F001-style feature IDs. Spawns the feature-evaluator agent in worktree isolation.