TimTGelhard
UserAdvanced operator's manual for Claude Code — principles, workflow discipline, per-project templates, and an opinionated global ~/.claude/ setup.
Categories
Indexed Skills (21)
op-add-skill
Use to add a new skill to the user's personal `bucket/skills/` library. Fires when the user says "save this as a skill," "make this a skill," "I want a skill for X," "add this to my bucket," "remember how to do this next time" (a recurring, reusable *procedure* — for a lightweight one-off note to revisit later, that's `/bucket suggest` / op-suggest, not this), or invokes `/bucket add`. Walks the user through naming, the trigger description, and the body; writes the skill file under `~/.achron-spine/bucket/skills/`; appends a row to `~/.achron-spine/bucket/INDEX.md`. Refuses if the pattern isn't reach-for-it-3+-times-worthy — see chapter 13's library thesis.
op-anti-patterns
Use BEFORE taking actions that match a known anti-pattern — about to add a new dependency, new abstraction, new helper "for future use", new chapter/skill/agent speculatively, "let me also add X while I'm here", "I'll write the docs after MVP works", merging a diff you haven't read, declaring UI done without browser-verifying, skipping a two-session RLS check, or extending the manual / skills / agents / hooks themselves. Also use when the user proposes one of those — this routes the anti-pattern *catalog*; the proactive reviewer-mode *cadence* for the same proposal is op-signaling (11e meta-scope), and both intentionally fire. Routes to chapter 18 (anti-patterns) of Achron Spine.
op-approach
Use BEFORE starting any non-trivial multi-domain or multi-session work. Fires on phrases like "audit X", "review X", "refactor X", "migrate X", "investigate X", "clean up X", "improve the project", "tackle this big thing", "the project has grown", "where do I start", "how should we approach", "step by step over multiple sessions". Identifies the work shape (Build / Audit / Refactor / Migration / Investigation / Research / Cleanup) and surfaces its default phase structure, hard rule, and common traps from the catalog in `chapters/workflow/05k-work-shapes.md`. ALWAYS fires before `op-prepare` for any work that isn't obviously a single-feature build. Skip for one-line edits or single-file changes — overhead exceeds payoff.
op-bucket-router
Use as the fallback router when the user's task isn't covered by any other core `op-*` skill — checks the user's personal bucket library (`~/.achron-spine/bucket/INDEX.md`, which has both a Skills table and a Chapters table) for a matching entry and loads only the matched file. Fires when another router would normally handle the task but no core chapter covers it (project-specific recipe, stack-specific pattern, user's personal convention, reference for their own architecture), or the user says "do this the way I usually do" / "use my X skill" / "check my bucket" / "is there anything in my chapters about this?". Never invents bucket entries — only routes to existing ones — and never fires on the `/bucket curate` or `/bucket suggest` slash commands (those are op-curate / op-suggest).
op-curate-nudge
When the bucket loop is enabled (default-off since round 6), auto-fires once at the start of any conversation if the user's bucket has 5+ pending suggestions AND it's been >30 days since the last `/bucket curate` run (or no `/bucket curate` has ever run). Emits one quiet line at the top of the first turn suggesting `/bucket curate`, then continues with the user's request. Silent otherwise — do not load the body when the conditions don't match.
op-curate
Use to run a curation session — review pending entries in `~/.achron-spine/bucket/SUGGESTIONS.md` and turn approved ones into new or modified files under `bucket/`. Fires on (1) `/bucket curate` slash command (with optional `--review-stale` for stale-entry review instead of pending curation); (2) the user explicitly asks to curate the bucket — "let's curate," "go through pending suggestions," "review what we captured," "process the queue," "let's review stale bucket entries." Shows a unified diff or full file body before every write — explicit per-change approval, no batch. Hard-refuses to touch `chapters/`, `skills/core/`, the global stub, or the profile — bucket-only.
op-foundations
Use when asking how Claude Code actually works under the hood, how the LLM loop and tools combine, what the three levers (context / scope / verification) are, why quality drops mid-session (drift / dilution / hallucination), how the 1M context window degrades before it fills, when to start a fresh terminal, what Claude genuinely can't do, which project types fit, which model (Opus / Sonnet / Haiku) to pick, when plan mode or fast mode is worth it, or how plan budgets and context cost work. To *recover* a session that's already degrading mid-flight (rather than understand why a failure mode happens), that's op-recovery. Routes to the foundations chapters (01–04) of Achron Spine.
op-hooks
Use when wiring an automatic, scripted behavior to a Claude Code event (typecheck after edit, block commit on test fail, notify on long completion), editing settings.json (global or project), troubleshooting why a hook didn't fire, or setting up the starter automation set for a new machine. For the hook-vs-CLAUDE.md-vs-skill decision ("from now on, when X, do Y" — where should it live?), op-persistence 12a is the front door. Routes to chapter 14 (hooks and automation) of Achron Spine.
op-onboard
Use to create or refresh `~/.claude/achron-spine-profile.md` — the personal calibration file capturing the user's Claude subscription, experience level, stack preferences, environment (OS / VCS / plans-dir / currency), project context (typical work / artifact / deploy target / database / team / users / org), working style, output format, and risk tolerance. Fires when the user invokes `/onboard` (essentials), `/onboard --deep` (full ~28-question interview, +2 conditional follow-ups for UI apps), or says "re-onboard" / "update my profile" / "redo onboarding". For a single output-style change, `/explain` (explanation style) or `/effort` (reasoning effort) is the lightweight path — this skill is the full profile interview. For the first-run greeting that points the user at `/onboard`, see the sibling `op-welcome` skill — it owns the file-absence surface so this skill never auto-launches an interview unannounced.
op-persistence
Use when deciding *where* a behavior or rule should persist across sessions — CLAUDE.md vs custom skill vs memory vs project doc vs hook. Fires on phrases like "should this go in a skill or CLAUDE.md?", "where do I save this rule?", "I keep telling Claude X every session — where should it live?", or when authoring a CLAUDE.md, drafting a skill or its trigger description, fixing a skill that isn't firing, or auditing a stale skill library. NOT for code-level persistence (localStorage, Redis, database schemas, session state). Routes to chapters 12 and 13 of Achron Spine; body branches into the sub-tasks once routed.
op-prepare
Use when starting a new project from scratch, scoping a major new capability in an existing project (>2 sessions of work), or whenever the user describes a multi-session piece of work and no project plan exists yet. Fires on phrases like "I want to build X", "let's start a new project", "let's plan the next big feature", "scope this out", "I have an idea for...", or when /prep is invoked. Walks the planning pass — brief → architecture → master plan → first section plan. Routes to chapters 05h (planning hierarchy), 05i (plan anatomy), and 05j (cold-start protocol) of Achron Spine.
op-recovery
Use when workflow quality is dropping mid-session, Claude is contradicting itself, hallucinating file paths or API names, drifting from earlier constraints, when each fix introduces a new bug, when the user says "are you sure", "you said the opposite earlier", "this isn't working", "you're going in circles", "I don't recognize this code", when a deploy broke production, a migration failed mid-flight, secrets leaked, or auth/RLS may be exposing user data. Routes to chapter 17 (recovery playbook) of Achron Spine.
op-suggest
Use to capture a high-signal moment to the user's personal queue at `~/.achron-spine/bucket/SUGGESTIONS.md`. Fires on four narrow conditions only — (1) explicit user signal ("we should add this to the manual," "remember this," "next time we hit this, let's…," "let's not forget this"); (2) repeated friction — Claude was corrected on the same pattern 2+ times in this session; (3) end-of-session reflection ("what did we learn here?," "anything worth capturing?," natural stopping point with a learning frame); (4) the user invokes `/bucket suggest`. Never fires on speculation ("it'd be cool if…"), one-off friction, mid-task ideation, or Claude's own hunches without user reaction. Appends one entry to the queue and gets out of the way — never modifies `chapters/` or `skills/core/`, never edits existing bucket files, never interrupts the task in progress.
op-tools
Use when choosing which Claude Code tool fits a task (Read/Edit/Write, Bash, grep, Agent, plan mode, TaskCreate, WebFetch, MCP, slash commands), debugging "should this be Bash or a dedicated tool", deciding when to background a long-running command, picking between grep and Agent(Explore) for a search, or auditing loaded MCPs for cost. Routes to chapter 15 (tool palette) of Achron Spine.
op-welcome
Auto-fires once at the start of any conversation when `~/.claude/achron-spine-profile.md` does NOT exist (a fresh install with no profile yet). Emits a short welcome block introducing Achron Spine and pointing the user at `/onboard` (the ≈3-min essentials interview) so the discovery surface is not "go read the README." If the profile file exists, this skill is silent — do not load its body. Never re-emits within the same conversation.
op-workflow
Use when starting a new project from scratch, scoping what to build before opening Claude Code, planning the prep / architecture / build / harden / ship sequence, sizing a feature for one session, choosing how to break a too-big feature, or deciding whether the current scope fits in one terminal. For the planning *pass* that actually writes the plan files (brief → architecture → master plan → first section plan), use op-prepare; this skill teaches the 7-stage *concepts* and feature sizing, not the plan-authoring pass. Routes to chapters 05 (the 7-stage workflow) and 06 (feature sizing) of Achron Spine.
op-brownfield
Use when entering a codebase you didn't build, returning to your own project after months away, inheriting client code, extending a site months after launch, or about to modify code where you couldn't answer "what does this file do" in one sentence. Covers the discovery sequence, safe first changes, teaching Claude an unfamiliar codebase, and when a rewrite is the right answer. Routes to chapter 08 (brownfield) of Achron Spine.
op-collaboration-modes
Use when deciding how to engage Claude for a task — whether to just say "build this" (executor), ask for an independent audit (reviewer), ask Claude to explain unfamiliar code before changing it (explainer), or get 2-3 design options with tradeoffs (planner). Also use when shifting modes mid-session (plan → build → review). For whether the *work itself* is audit-shaped or multi-session in shape (vs which engagement mode to take), that's op-approach. Routes to chapter 07 (collaboration modes) of Achron Spine.
op-prompting
Use when writing or improving a prompt, deciding how to phrase a request to Claude Code, structuring a non-trivial task (CONTEXT/TASK/CONSTRAINTS/EXAMPLES/OUTPUT), iterating after a wrong first output, or picking a high-leverage prompt pattern (orientation, challenge-me, small-fix, what's-broken, explain-it-to-me). Routes to chapter 09 (prompting) of Achron Spine.
op-subagents
Use when deciding whether to delegate to a subagent (Agent tool) vs do work inline, writing a subagent prompt, picking an agent type (general-purpose / Explore / Plan / custom), running parallel or background agents, designing a custom subagent in ~/.claude/agents/, or evaluating whether an orchestrator-with-specialists pattern fits the task. Routes to chapter 16 (subagents) of Achron Spine.
op-visuals
Use when deciding whether to paste a screenshot vs text, sharing a UI bug or design reference with Claude, writing ASCII or Mermaid diagrams in project docs, sketching a UI mockup before a build session, or handling mobile / cross-platform visual comparisons. Routes to chapter 10 (visuals) of Achron Spine.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.