/lazy-core.help

Output the block below verbatim to the user. Do not summarize, rephrase, or add commentary. Do not invoke any tools. Do not log this run.

---

lazycortex-core — opinionated hygiene layer for Claude Code configs. Audits what's loading into context, slims oversized rules, flags secrets/PII/paths before public commits, and batches MCP tool permissions.

Skills (invoke as /<name> or via Skill tool):

• lazy-core.agent-models — interactively assign model tiers (haiku/sonnet/opus/inherit) to every dispatchable subagent missing from lazy.settings.json; routes each entry to its structurally-correct scope.
• lazy-core.audit — read-only report of what gets loaded into context at startup, by category and size.
• lazy-core.doctor — health check across rules, agents, skills, commands, settings, memory, hooks, CLAUDE.md. Delegates to sibling audits.
• lazy-core.install — bootstrap the plugin into a project (or globally); copies hygiene + security rule templates. Idempotent.
• lazy-core.optimize — slim oversized rule files, move reference material to agents, fix global-vs-local settings leakage.
• lazy-core.setup — meta-installer that runs every enabled lazycortex plugin's install skill in dependency order. Idempotent.
• lazy-expert.cancel-job — cancel an expert job by removing its directory. Confirms via AskUserQuestion for non-done jobs.
• lazy-expert.collect-job — collect the result of a dispatched expert job; returns {status, response}.
• lazy-expert.dispatch-job — dispatch a job to a named expert queue; returns {job_id, queue_path}.
• lazy-expert.list-jobs — list expert queue jobs, optionally filtered by expert name or status.
• lazy-guard.allow-mcp — add all tools of one or more MCP servers to permissions.allow at the correct scope.
• lazy-guard.check-public — scan a repo (or subtree) for leaked secrets, PII, internal hostnames, hardcoded local paths. Respects .guard-waivers.json.
• lazy-repo.mark-public — walk a repo through the check-public audit, create a waivers file, optionally flip GitHub visibility.
• lazy-log.clean — interactive housekeeping for .logs/claude/; classifies each run-log directory against the live skill/agent/command name set, surfaces orphans and unnamed clusters, and applies merge/delete/leave per user confirmation.
• lazy-routine.register — register a named routine in lazy-core.runtime; used by plugin install skills.
• lazy-routine.unregister — remove a named routine from lazy-core.runtime; protects the built-in expert-pump.

Commands (invoke as /<name>):

• lazy-core.checkup — single entry point: runs every read-only audit/doctor (consumer + author trios), merges findings into one per-plugin table, then prompts for the mutating fix-flow to run.
• lazy-core.help — this listing.

Agents (dispatched via Agent(subagent_type: "lazycortex-core:<name>")):

• lazy-log.bullets — drafts user-facing changelog bullet blocks from recent distilled entries.
• lazy-log.distill — rolls commits in .logs/commits.jsonl into themed changelog entries in .logs/changelog.md.
• lazy-log.recall — searches change history across .logs/changelog.md, per-run log files, .logs/commits.jsonl, and git log; returns ranked results with git SHAs.
• lazy-log.summary — synthesises a multi-source narrative summary for a given topic or time range.
• lazy-log.timeline — produces a chronological view of changes for a date range or topic.

Rule:

• lazy-log.logging — run-logging contract; every skill, agent, and command must log each run to .logs/claude/<name>/YYYY-MM-DD_HH-MM-SS.md.

Documentation:

• agent-models — Assign haiku/sonnet/opus tiers to every agent in your vault and let the model-router hook route each dispatch automatically.
• change-history — Run-log housekeeping and change-history access — clean up orphaned log directories, distill commits into themed prose, and ask "why was X changed?" across every source at once.
• experts — Dispatch jobs to named expert workers, keep the main session free, and collect results when the daemon finishes them.
• git-coordination — Inspect and manually break the per-repo staging lock that prevents concurrent Claude Code sessions from corrupting each other's git index changes.
• guardian — Catch secrets, PII, and internal paths before they reach a public repo; stop per-tool allow prompts for new MCP servers in one step.
• install-and-audit — Bootstrap and verify lazycortex-core — the shared scaffolding layer every other plugin depends on.
• memory — Per-expert long-term memory tracked in git — experts consult notes before primary work, write new notes as a side-effect of jobs, and consolidate via reflect passes.
• runtime — Register, unregister, and recover routines in the per-repo serial daemon — five routine types keep the async team running in order; the recovery skill handles both dirty-tree and remote-sync halts.
• add-memory-to-expert — Opt an existing expert into the memory subsystem, dispatch jobs to accumulate runs, run the first reflect pass, and verify the expert's first durable notes land in .memory/.
• make-repo-public — Step-by-step guide to making a repo public safely — audit, fix secrets, set your public author identity, create the waiver file, and flip GitHub visibility.
• setup-expert — Add a named expert role and dispatch your first async job — keep working while the daemon runs it, then collect the result.
• setup-routine — Register a dot-namespaced periodic routine with the runtime daemon and remove it cleanly when it is no longer needed.
• setup-runtime — Bootstrap the per-repo serial daemon so the async expert team has an executor — install wizard, start the daemon, then unblock it with /lazy-runtime.recover if the working tree halts.
• troubleshooting — Common failure modes across lazycortex-core skills — symptoms, likely causes, and fixes.
• faq — Answers to non-obvious questions about skill selection, upgrade flows, settings placement, plugin composition, agent routing, MCP scope decisions, the expert runtime, memory subsystem, routine types, git coordination, change-history agents, and run-log housekeeping.

Offline copy at ~/.claude/plugins/cache/.../claude/lazycortex-core/help/.

See README.md in the plugin for full scenarios and examples.