By ygavrilov
A personal operating system for Claude Code: a PA agent that classifies every input and delegates to worker tiers running SOP skills for task, wiki, planning, scheduling, notes, and software-delivery workflows. Ships generic — a first-run setup wizard personalizes your workspace.
Junior Worker — fast low-cost executor. Process-agnostic. Executes a single delegated task by invoking the relevant SOP skill. Use for simple, well-defined, low-stakes tasks. Spawned by PA, never self-directed.
Lead Worker — highest-capability executor. Process-agnostic. Executes a single delegated task by invoking the relevant SOP skill. Use for complex or high-stakes tasks requiring strongest reasoning. Spawned by PA, never self-directed.
Personal Assistant — single point of contact for all inputs. Receives any message, classifies it, and routes to the correct handler (calendar, wiki, inbox, drafts, or direct execution). Use proactively for any input that needs to be processed, scheduled, captured, stored, or replied to.
Senior Worker — mid-capability executor. Process-agnostic. Executes a single delegated task by invoking the relevant SOP skill. Use for standard tasks of moderate complexity. Spawned by PA, never self-directed.
SOP to run verified, on-demand research on any question and write a structured research report to research/. Adapted from the ICOR "Pax" research-agent protocol — triangulate sources, mark confidence, lead with the actionable answer. Use when a worker or PA is asked to research, investigate, fact-check, compare options, or brief before a decision. Never writes the wiki directly — flags durable findings instead.
Scan every project skill's allowed-tools frontmatter and write a merged, deduplicated, sorted allowlist into the project .claude/settings.json, so all skill tool permissions are enabled by default with no per-session prompts. Use when the user says "setup skill permissions", "regenerate permissions", "update settings allowlist", or invokes /setup-skill-permissions. Re-run after any skill is created or its allowed-tools changes.
SOP to design a technical solution and produce an Architecture Decision Document (ADD) — context review, options analysis (2-3 approaches with trade-offs), per-layer deep-dive (DB / backend / frontend / infra / SEO), risks, and decision. Use when a worker is delegated a dev architecture task, or asked to "design a solution", "produce an ADD", "evaluate technical options", "decide the architecture". SDLC Phase 2. Best run by the lead tier.
SOP to implement a single Work Breakdown task / vertical slice against a repo — read context, locate the code, make the change following the project's existing conventions, run the build/tests locally, and update PROGRESS.md. Use when a worker is delegated a dev implementation task, or asked to "implement this slice", "build this feature task", "write the code for X". SDLC Phase 5 (per-task execution). Tier by task complexity. Stack-agnostic.
SOP to onboard an existing codebase into personal-os — analyze a repo across technology / architecture / standards / state and produce a populated PROJECT_CONTEXT.md in the target project's .claude/. Use when a worker is delegated a dev task against a project that has no PROJECT_CONTEXT.md yet, or asked to "ingest a repo", "onboard a codebase", "analyze this project". Mandatory prerequisite before any other dev SOP runs against a repo.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A personal operating system for Claude Code. A PA (Personal Assistant) agent is your single point of contact: it classifies every input and delegates to a worker tier running a leaf SOP skill. Out of the box it ships generic — zero personal data baked in — and a first-run wizard personalizes your workspace.
pa plus three worker tiers (junior-worker, senior-worker, lead-worker).sop-task, sop-wiki, sop-plan, sop-schedule, sop-draft-reply, research, setup-skill-permissions, plus the knowledge/* sub-skills.workflow-setup-os (the setup wizard), workflow-process-message, workflow-ingest-notes, workflow-triage-tasks.sop-dev-ingest-repo, sop-dev-requirements, sop-dev-architecture, sop-dev-work-breakdown, sop-dev-implement, sop-dev-test, and the workflow-deliver-feature SDLC orchestrator. The dev SOPs are stack-agnostic — they read the target repo's PROJECT_CONTEXT.md and follow its existing conventions.CLAUDE.md (with a ## Config section), framework.md, wiki SCHEMA.md, and the task template.sop-schedule,
sop-plan) can create real calendar events when a calendar MCP (e.g.
mcp-ical) is available. It is
optional: if you set mcp_ical_present = n in your Config, scheduling
degrades gracefully — the skill still recommends a slot and records it in
the task/plan, but does not create an event. Configure your calendar MCP in
your environment's .mcp.json if you want live events.From a marketplace:
/plugin marketplace add ygavrilov/plugin-personal-os
/plugin install personal-os@personal-os-marketplace
Or load locally for testing:
claude --plugin-dir /path/to/plugin-personal-os
After install, run /reload-plugins (or restart) so the skills and agents register.
personal-os:workflow-setup-os (or say "set up personal os").mcp_ical_present (y/n), primary calendar, read-only reference calendars.## Config section of CLAUDE.md,
lays down framework.md from template, scaffolds your workspace
(wiki/, tasks/, weeks/, drafts/, research/, inbox.md), and
regenerates skill permissions.The wizard is idempotent — re-run it any time to change a group; it never destroys your data.
Personal OS scores work by self-benefit using dimensions and weights you
choose. There are no defaults. Run sop-plan (quarterly or weekly) for your
first planning session — that's where you define your dimensions and their
weights. They are recorded in the ## Config section of CLAUDE.md, which
framework.md treats as the authoritative source.
On install, the plugin makes the PA agent the active orchestrator and enforces the route-everything-through-PA workflow. Two supported plugin mechanisms do this:
settings.json containing
{ "agent": "pa" }. Claude Code applies this when the plugin is enabled,
activating the PA agent as the main thread (its system prompt, tools, and
model) — so PA is the single point of contact without any manual selection.UserPromptSubmit hook
(hooks/hooks.json → hooks/enforce-workflow.js, resolved via
${CLAUDE_PLUGIN_ROOT}) injects a MANDATORY PA PROTOCOL into context on
every prompt: before spawning any worker/Agent, PA must invoke
workflow-process-message and route the input. Build, config, and meta
requests are not exempt.Together these make installing the plugin enough to have PA drive the workspace.
After install run /reload-plugins (or restart) so the agent setting and hook
register. The consuming workspace's CLAUDE.md (from the template) also states
PA is the single entry point, reinforcing the behavior for any session.
npx claudepluginhub ygavrilov/plugin-personal-os --plugin personal-osDelivery Manager plugin v3 — orchestrator + engineer workers + core/lang/framework/project skills.
A plugin providing essential skills for creating agents, skills and managing llm wiki. While agents and skills are based on the official documentation they are highly opinionated.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Superpowers Plus core skills library for Claude Code: planning, execution routing, TDD, debugging, and collaboration workflows
Comprehensive C4 architecture documentation workflow with bottom-up code analysis, component synthesis, container mapping, and context diagram generation
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
v9.44.1 — Patch release for Gemini environment/version detection and qwen auth gating. Run /octo:setup.