Development

What the gates cannot tell you about adding or changing a zh-TW rule - why assets/ruleset.json is the only place a vocabulary rule lives, the false-friend problem and the four gates that answer it, the corpus thresholds a rule has to clear, why positions are byte offsets through NFC, and how a rule reaches the scanner, the fixer and the browser build. Use when adding or disabling a rule, when a rule fires on native zh-TW prose, when a fix lands at the wrong offset, or when touching src/engine...

Use when implementing any feature or bugfix, before writing implementation code

Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.

Anti-slop frontend skill for landing pages, portfolios, and redesigns. The agent reads the brief, infers the right design direction, and ships interfaces that do not look templated. Real design systems when applicable, audit-first on redesigns, strict pre-flight check.

Write what GitHub itself reads: README, CONTRIBUTING, the community-health files (SECURITY, CODE_OF_CONDUCT, LICENSE, issue and PR templates) and the workflows under .github/workflows/. Use when creating or rewriting any of these, when Community Standards reports a gap, when adding or reviewing CI, or when a repository goes public. Not for docs/, which is writing-docs.

Find a Claude Code skill or plugin somebody else has already written, and judge whether to install it. Use before writing a skill for something that sounds common, when asked whether an integration exists, when looking for a plugin by capability rather than name, or before adding a skill that would duplicate one installed.

Firmware development for the Guition JC-ESP32P4-M3-DEV board (JC-ESP32P4-M3 module — ESP32-P4NRW32 with 32 MB in-package PSRAM, 16 MB flash, and an ESP32-C6 radio co-processor on SDIO) — its MIPI-DSI panel and MIPI-CSI camera connectors, IP101 Ethernet, ES8311 codec with speaker amplifier and microphone, microSD on SDMMC, RS485, three USB-C sockets, the 26-pin JP1 header, and the PlatformIO + ESP-IDF (pioarduino) setup around them. Use when working on this board or any ESP32-P4 devkit: projec...

Create or revise living Designs that reconcile required behavior, technical realization, decisions and acceptance intent. Use for living-Design authorship and source reconciliation before independent Design Review.

Model validation contract: model-document-v1 For this repository’s [complete source retirement](../../changes/2026-09-14-retire-specs-and-stale-tests/source-disposition.md), current responsibilities are self-contained in the owning Designs. Original source-transfer inventories remain recoverable through [Historical provenance](#historical-provenance); their instructions to retain or amend legacy specs, architecture, activation state or retired engines are historical and superseded by this com...

Repository または subsystem を対象に、source-only で明示的なセキュリティ監査を行う entry skill。 通常の PR セキュリティレビューとは分離し、reconnaissance、scope 固定、既存 security skill への委譲、 evidence と unresolved hypothesis、observe-only SecurityAuditCoverage、coverage critic、 structured audit artifact と repeat-run coverage を扱う。target-controlled code は実行しない。

Notes for `docs/research/118-phase-256-the-architecture-that-explains-itself.md`. Everything below was read at `/Users/gdc/as-built-architecture` (read only, never written, `git log` head `aa0428d`) and measured by RUNNING the helpers over scratch copies. Line numbers are of the files as they stand at that head.

Audit interface source with HIG Doctor when the user asks to scan, lint, review, or remediate Swift, UIKit, AppKit, React, Vue, Svelte, Angular, Compose, Android XML, Flutter, CSS, or HTML for HIG-aligned interface and accessibility concerns. Do not activate for general design advice without source to audit; use the relevant HIG guidance skill instead.

Create, edit, review, arm, and debug vincent event triggers, the YAML under {config_dir}/triggers and the poll scripts they run. Use for trigger sources (command, github_issues, github_prs, http, schedule), match and if filters, dedupe keys, limits, on_fire, permission, arming, dry runs, the delivery ledger, or trigger validation errors. Do not use for vincent workflows, including the workflow a trigger's action.workflow names (use vincent-workflows), or for GitHub Actions.

Implement a piece of work based on a spec or set of tickets.

Ask which skill or flow fits your situation. A router over the skills in this repo.

The SOFTWARE domain profile for the strategist. A domain profile answers the five questions the strategist's method deliberately leaves open: what surveying the current state means, what two concurrent tickets contend for, which capabilities the work routes to, what sections a plan carries, and what sections a ticket carries. This profile answers all five for a code project — survey the codebase, contend over file paths, route to developer/test-writer/reviewer, and carry Codebase Analysis and...

Teach decision-seat holders the CLI contract through a bundled system skill. Select that skill from the seat-derived advisory action and make the review and approval prompts require the command as their final action.

Record the current workflow step decision through the task-bound Office CLI.

Author a Kandev task canvas as a self-contained web application.

Create branded architecture, IT current-state, flowchart, sequence, state machine, ER/data model, timeline, swimlane, quadrant, radar/spider, polar chart (polar/radial lollipop), loop/flywheel, nested, tree, org chart, layer stack, Venn, pyramid/funnel, treemap, bar, line, Gantt and scatter charts, high-level, process, medallion, data flow, DP integration, DP security matrix, Sankey, fishbone, Wardley map, kanban, user journey, deployment, dependency graph, UML class, story map, or database s...

Print your own north-star row from your own record, and nothing else. Reads .claude/harness/decision-log.md, the canvas and the hook logs in this project; prints one line (decisions before the first source file, of which citing outside evidence, kills before code, sessions). Sends nothing anywhere. Opt-in; paste the line where you like or not at all.

Write or edit a Session Sitter practices file — the markdown that decides permission prompts. Use when the user wants to add a rule, block a command, allow something without approval, set up a practices file, or asks why a clause did not fire.

Read and interpret the Session Sitter audit trail — what agents were allowed to do, which clause was applied, who decided, and how long it took. Use when the user asks what happened overnight, why a call was denied or rewritten, which rules are firing, or wants to export the decision log.

Find out whether Session Sitter will allow a tool call, and which written clause decides it, BEFORE running it. Use when about to run something plausibly governed — a push, a deploy, a migration, anything touching secrets or infrastructure — or when the user asks what the policy allows, or why a call was denied.