miklschmidt/archboard / skills/archboard-dev
archboard-dev
VerifiedKeep Archboard's canonical architecture model, consumer guidance and docs site aligned with product changes. Use when planning, implementing, or closing work in this repository.
Work on Archboard
Use the consumer skill for all architecture operations and authoring rules.
Use the vault guide for this repository's context.
Discover authority through archboard model read in .archboard/vault.
Plan the change
Identify affected modules, relationships, flows, and walkthroughs, then record their IDs in task notes.
Read affected modules through model inspect and their change patterns through model locality.
If the architecture stays accurate, record why no proposal is needed.
Documentation of implemented architecture needs no second proposal.
Before implementation, create a compact proposal for each independent architectural intent through the consumer proposal recipe. Record stable state ID, predecessor, model, and vault in the task plan.
Finish the change
Update the proposal to describe implementation, including affected subjects absent from the plan.
Compare its state with the intended base through model compare.
Compare the branch with the proposal through model footprint.
Settle affected proposal and descendant standing through the consumer reconciliation recipe.
Run focused semantic and code inspection for the affected state. Inspect source for claims that automated evidence cannot assess. Fix findings caused by the task, including containment findings on touched subjects from the consumer authoring reference. Record unavailable source and evidence limits in task notes. Use the narrow validation owners required by the behavior. Reserve the complete repository gate for integration acceptance or an explicit request.
Follow consumer completion, retention, publication, and recovery recipes for adoption, shelving, and shared evidence. Commit affected authored inputs with the work.
Keep the docs in sync
The docs site in site/ describes the shipped CLI and the GitHub Actions.
Sync it in the same change when the change alters one of these:
- A command, option, example, output, exit code or refusal of the CLI.
- A finding or skip reason of a check, or a model schema.
- An input or the behavior of a GitHub Action.
- The install or release steps.
Generated pages come from the program: site/scripts/reference.ts reads the CLI contracts, the help of each check, the JSON Schemas and each actions/*/action.yml.
To change what a generated page says, edit its source in the CLI or the Action. The page only lays out that data.
The site's prebuild and predev run the script. The script throws when an example runs another command, when the embeds help misses a code or the code help misses a skip reason, or when a description has a paragraph over 120 words or 6 sentences.
Write a description in the format that CommandContract.description documents: paragraphs, colon heads over lists, and code in backticks.
Hand-written pages quote behavior: site/src/content/docs/*.mdx, the prose of the pages in site/src/pages/docs/, actions/README.md, INSTALL.md and README.md.
A new top-level command needs a page in cliGroups of site/src/lib/docs-nav.ts. The build refuses a command that no page names.
Search them for each changed command, flag, code and exit status, and correct every statement and example that the change makes false.
The build guards the guides with two more scripts:
site/scripts/guides.tsrefuses anarchboardcommand or flag in code that the CLI lacks, and awith:key that an Action lacks. It also refuses a code span that names an Action input or output, or a finding code, without a link to its row in the reference. It refuses an exit code written by hand, which<Exit>reads from the contract, an Action value or default that theaction.ymldoes not give, a workflow whose Node.js differs from the check Action's default, a permission that an Action'saction.ymldoes not name, a workflow without thefetch-depth: 0that an Action asks for, and anARCHBOARD_…variable that the CLI does not read.site/scripts/links.tsrefuses a dead internal link or anchor indist, and a guide's{reference.…}expression that shows as text. So a renamed input or code breaks the build at each guide that links it.
A guide reads a number of the program from reference.facts, which reference.ts imports from the source, such as the release port. Add a fact there when a guide states a new one.
Each release has its own docs (ADR 0077). A change to the site reaches /docs/next/ at once, and /docs/ at the next release. The release workflow builds the docs of the tag and attaches them to the GitHub release, so a release needs no docs step. In a guide, write {{version}} where it pins the CLI and @{{sha}} # {{ref}} where it pins an Action. The build refuses a version or an Action ref that a guide pins by hand.
Follow site/DESIGN.md for anything a reader sees.
Run nix develop -c bun run build in site/, then read each changed page in the running site with bunx agent-browser.
Finish when the build passes and no page describes behavior that the change replaced.
Maintain consumer guidance
skills/archboard serves agents in other repositories.
When product behavior changes agent actions, update affected guidance in the same change.
Derive commands from shipped CLI contracts and schemas.
Keep general guidance independent of repository and language. Put repository-specific detail in worked examples.
When evaluated behavior changes, update affected scenarios and coverage under evals/, outside the distributed skill.
Use evaluation procedures for deterministic validation.
Leave model evaluations to the person who runs them.
Measure workflow completion and total tokens, including discovery and recovery. Compare runs only under the same product contract. If that contract changes, establish a new baseline before claiming token savings.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 469k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 197k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 197k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 197k | Updated | Claude CodeCodex | Verified | Install |