toolsmith
When you notice you are doing the same thing BY HAND again — the same check, trace, scaffold, or detection you did before — that noticing is the signal for this. It is the instinct to externalize repeated effort into tools: never do the same manual work twice. Use this when a motion has recurred: "I keep hand-checking whether effects leaked into the core", "I keep manually tracing the call graph". The Rule of Three pointed at Cairn's OWN repeated actions: the third time you do a thing by hand, forge a tool so the fourth is free. Tools live in a sandboxed workshop and may take real actions (run commands, scaffold, probe), trusting the outer harness (Claude Code) to bound the blast radius — a tool the harness rejects is itself a lesson. A tool is justified by MEASURED payback (manual cost x times done, vs build cost); one that never pays back is debt, and is retired. The bright line: a tool may never rewrite Cairn's own constitution — gates, ledger, floor, ladder.
What this skill does
# Toolsmith — never do the same manual work twice The most basic thing a builder does: when the work is repetitive, make a thing so it isn't. Cairn can learn, recall, grow trust, and think under fog — but without this it works with bare hands forever. This is the faculty that makes the others compound, because a tool doesn't just help once; it changes what is CHEAP, and changing what is cheap changes what Cairn will attempt. ## Why this is the third way the cost curve bends The marginal cost of feature N+1 slopes down three ways: the shelf (reusable code parts), knowledge (don't re-derive), and — the most powerful because it compounds — reusable CAPABILITY: tools Cairn forges for its own working. A mental-model makes you think better next time; a tool means you need not think at all next time, because the motion is automated. Tools become substrate the next tool is built from. That is the runaway humans have and other learners don't. ## The trigger — Rule of Three on your OWN motions The ratchet already promotes a fix when a defect recurs and a shelf item when an abstraction recurs. This is the third recurrence it was blind to: a repeated MANUAL MOTION. The third time you catch yourself doing the same hand-work — a check, a trace, a scaffold, a probe — that is a tool waiting to be born. Two times might be coincidence; three is a pattern with a shape clear enough to automate correctly. Building on the first motion is how you forge the wrong tool, more expensive than the manual work it replaced. ## Tools may act — the harness is the safety, and rejection teaches A tool may take real actions: run a command, scaffold a file, probe the repo. Cairn does not need training wheels here, because it does not work alone — it runs inside an outer harness (Claude Code) with permission prompts, file boundaries, version control, and a human in the loop. A tool that tries something destructive hits that harness and is REJECTED, and that rejection is not a failure — it is the cheapest falsifying observation about the tool. Treat it as inquiry applied to your own creations: you predicted the tool was safe and useful; the harness disagreed; record the surprise, update, and rebuild. Cairn is not the last line of defense and must not act as if it were. ## The one bright line A tool may help Cairn WORK; a tool may never rewrite Cairn's own CONSTITUTION. The gates, the capability-ledger math, the floor semantics, the maturity ladder — these are the substrate made unchangeable-by-learning on purpose, and a tool that edits them is the agent-psychosis failure with a power drill. Tools live in the workshop (`.cairn/tools/`), sandboxed from the skills' own scripts and stores. A tool is a lever Cairn picks up, never a rewrite of Cairn's spine. ## Justified by measured payback, retired if it doesn't earn its keep A tool is not justified by being clever; it is justified by the manual cost it eliminates. Record: this motion cost ~N steps by hand, done K times; the tool cost M to build; it pays back after about M/N reuses. A tool that never reaches payback is debt, not leverage — the same law as premature abstraction, pointed at the workshop. The instinct to build is paired with the honesty to retire a tool that didn't earn its keep, so the workshop stays a toolbox and not a junk drawer. ## Discoverable, or it gets rebuilt A tool nobody can find is a tool that gets rebuilt, and you have grown duplication in the workshop instead of the product. Every tool is catalogued by the motion it replaces, so the next time the smell appears, the tool surfaces instead of being forged again. ## Files - `references/forging.md` — when a motion is ripe, how to scope a tool small, and how to compute payback. - `scripts/store.py` — the workshop catalogue (JSONL): tools by the motion they replace. - `scripts/motion_observe.py` — log a repeated manual motion; flags it RIPE at three. - `scripts/tool_forge.py` — register a forged tool with its payback math; refuses a tool whose target path is inside a skill or store (the bright line). - `scripts/tool_check.py` — surface the tool for a motion; report tools below payback as candidates to retire.
Related in AI Agents
skill-development
IncludedComprehensive meta-skill for creating, managing, validating, auditing, and distributing Claude Code skills and slash commands (unified in v2.1.3+). Provides skill templates, creation workflows, validation patterns, audit checklists, naming conventions, YAML frontmatter guidance, progressive disclosure examples, and best practices lookup. Use when creating new skills, validating existing skills, auditing skill quality, understanding skill architecture, needing skill templates, learning about YAML frontmatter requirements, progressive disclosure patterns, tool restrictions (allowed-tools), skill composition, skill naming conventions, troubleshooting skill activation issues, creating custom slash commands, configuring command frontmatter, using command arguments ($ARGUMENTS, $1, $2), bash execution in commands, file references in commands, command namespacing, plugin commands, MCP slash commands, Skill tool configuration, or deciding between skills vs slash commands. Delegates to docs-management skill for official documentation.
reprompter
IncludedTransform messy prompts into well-structured, effective prompts — single or multi-agent. Use when: "reprompt", "reprompt this", "clean up this prompt", "structure my prompt", rough text needing XML tags and best practices, "reprompter teams", "repromptception", "run with quality", "smart run", "smart agents", multi-agent tasks, audits, parallel work, anything going to agent teams. Don't use when: simple Q&A, pure chat, immediate execution-only tasks. See "Don't Use When" section for details. Outputs: Structured XML/Markdown prompt, quality score (before/after), optional team brief + per-agent sub-prompts, agent team output files. Success criteria: Single mode quality score ≥ 7/10; Repromptception per-agent prompt quality score 8+/10; all required sections present, actionable and specific.
adaptive-compaction
IncludedAdaptive add-on policy and recovery layer that decides WHEN to compact, prune, snapshot, or fork -- replacing fixed-percent auto-compaction across Claude Code, Codex, and MCP-capable hosts. Trigger on auto-compact timing or damage: "when should I compact", "is it safe to compact now or start a fresh session", "auto-compact fires too early/mid-task", "switching to an unrelated task but the window still has space", "context rot", "answers get worse the longer the session runs", "the agent forgot the plan or my decisions after it summarized", "add a layer on top that manages context without changing the agent", raising autoCompactWindow to give the policy room, or installing/tuning a cross-tool compaction policy or PreCompact hook -- even when "compaction" is never said but the problem is context-window pressure or post-summarization memory loss. Do NOT use to summarize a conversation, build RAG, write a summarization prompt (decides WHEN not HOW), or answer max-context-length trivia.
agent-skill-creator
IncludedCreate cross-platform agent skills from workflow descriptions. Activates when users ask to create an agent, automate a repetitive workflow, create a custom skill, or need advanced agent creation. Triggers on phrases like create agent for, automate workflow, create skill for, every day I have to, daily I need to, turn process into agent, need to automate, create a cross-platform skill, validate this skill, export this skill, migrate this skill. Supports single skills, multi-agent suites, transcript processing, template-based creation, interactive configuration, cross-platform export, and spec validation.
llm-wiki
IncludedUse when building or maintaining a persistent personal knowledge base (second brain) in Obsidian where an LLM incrementally ingests sources, updates entity/concept pages, maintains cross-references, and keeps a synthesis current. Triggers include "second brain", "Obsidian wiki", "personal knowledge management", "ingest this paper/article/book", "build a research wiki", "compound knowledge", "Memex", or whenever the user wants knowledge to accumulate across sessions instead of being re-derived by RAG on every query.
skill-master
IncludedAgent Skills authoring, evaluation, and optimization. Create, edit, validate, benchmark, and improve skills following the agentskills.io specification. Use when designing SKILL.md files, structuring skill folders (references, scripts, assets), ingesting external documentation into skills, running trigger evals, benchmarking skill quality, optimizing descriptions, or performing blind A/B comparisons. Keywords: agentskills.io, SKILL.md, skill authoring, eval, benchmark, trigger optimization.