scaffold

Plugin scaffold from marketplace ia-tools in julianjab/ia-tools. Commands read from the repo's manifest — 1 month ago.

Skills (12)

• audit-agent

  Use when the user asks to review or validate an existing Claude Code agent file (agents/<name>.md) against best practices. Reports frontmatter drops, first-person descriptions, description quality, tool-allowlist issues, model mismatches, and the 14 anti-patterns. Read-only; never edits the agent.

• audit-mcp

  Use when the user asks to review or validate an MCP server plugin (plugins/<name>/) against best practices. Checks directory layout, .mcp.json format, plugin.json, tool descriptions, Zod schemas, stdio hygiene, dist-in-sync, and test coverage. Read-only.

• audit-script

  Use when the user asks to review or validate a bash hook, repo script, or skill bash block against ia-tools structured-bash style rules. Reports missing doc headers, unquoted expansions, mixed-bucket responsibilities, unsafe piping, and the 20 script-style rules (S1-S20). Read-only; never edits the

• audit-skill

  Use when the user asks to review or validate an existing Claude Code skill (skills/<name>/SKILL.md) against best practices. Reports hardcoded paths, label-shaped descriptions, first-person descriptions, deep reference chains, missing argument tables, unscoped tool permissions, and the 18 skill anti-

• edit-agent

  Use when the user asks to modify an existing Claude Code agent definition (agents/<name>.md). Audits the file, delegates the change to the agent-author subagent in edit mode, then re-audits to confirm no regressions. Interactive — collects a change request before editing.

• edit-mcp

  Use when the user asks to modify an existing MCP server plugin (plugins/<name>/). Audits the plugin, delegates source-level changes to the mcp-author subagent in edit mode, then re-audits to confirm no regressions. Reads-only against manifests; the calling user owns plugin.json / .mcp.json / package

• edit-script

  Use when the user asks to modify an existing bash script (hook under hooks/scripts/, repo helper script, or a skill's bash block), or to auto-fix /audit-script findings against script-style.md. Bracketed by /audit-script runs before and after so regressions surface explicitly.

• edit-skill

  Use when the user asks to modify an existing Claude Code skill (skills/<name>/SKILL.md or sibling files). Audits the directory, delegates the change to the skill-author subagent in edit mode, then re-audits to confirm no regressions. Interactive — collects a change request before editing.

• new-agent

  Use when the user asks to create a new Claude Code agent. Gathers a brief, delegates design to the agent-author subagent, writes the agents/<name>.md file, then runs /audit-agent on it. Interactive — asks clarifying questions before writing.

• new-mcp

  Use when the user asks to create a new MCP server plugin. Runs the deterministic scaffold.sh with templates to produce plugins/<name>/ — plugin.json, .mcp.json, package.json, tsconfig, src/mcp-server.ts, tests. Then runs /audit-mcp. Optionally delegates custom tool design to the mcp-author subagent.

• new-script

  Use when the user asks to create a new bash hook, repo script, or skill bash block following ia-tools structured-bash style. Gathers a brief, delegates design to the script-author subagent, writes the file, then runs /audit-script on it.

• new-skill

  Use when the user asks to create a new Claude Code skill (slash command). Gathers a brief, delegates design to the skill-author subagent, writes the skills/<name>/ directory, then runs /audit-skill. Interactive — asks clarifying questions before writing.

Subagents (4)

• agent-author

  Use when the user asks to design, generate, or modify a Claude Code agent definition. Produces a complete agents/<name>.md following the ia-tools best-practice references. Supports mode=new (write fresh file) and mode=edit (apply a focused change to an existing file). Reviews its own output against

• mcp-author

  Use when /new-mcp --custom or /edit-mcp delegates MCP source design. Customizes src/mcp-server.ts, src/shared/types.ts, src/__tests__/server.test.ts (and optional src/tools/*.ts) inside an already-scaffolded plugin directory. Supports mode=new (post-scaffold customization) and mode=edit (focused cha

• script-author

  Use when the user asks to design, generate, or modify a bash script — a hook under plugins/<plugin>/hooks/scripts/<bucket>/, a repo script under scripts/, or a skill bash block. Produces a complete .sh file that passes the structured-bash style rules S1–S20, and optionally patches plugins/<plugin>/h

• skill-author

  Use when the user asks to design, generate, or modify a Claude Code skill (slash command). Produces or updates a skills/<name>/ directory (SKILL.md + sibling scripts/templates). Supports mode=new (create directory) and mode=edit (apply a focused change to an existing skill). Reviews its own output a

Hooks (1)

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/audit-on-definition-edit.sh

Similar plugins