Skip to content
agent-skills-toolkit

askit-build-command

Reference: askit-build-command

Section titled “Reference: askit-build-command”

Creates and improves Claude slash commands (commands/<name>.md) to the Standard. Claude-native; no Codex artifact.

  • create: brief interview (command name, backing skill or workflow, optional argument hint, description) -> copy templates/command.md to commands/<name>.md -> fill frontmatter (description, maps-to, optional argument-hint/ allowed-tools/model, metadata.version) -> write a body invoking the backing skill via $ARGUMENTS -> declare in library.json components.commands -> run node scripts/evaluate.mjs . --json to 0 errors (S3 + S7 must be clean).
  • improve: consume node scripts/evaluate.mjs . --json -> resolve findings (S3 index drift, S7 missing maps-to/unresolved target/empty description) -> re-evaluate to 0 errors.
  • Files touched: commands/<name>.md (created or edited), library.json (components.commands entry).
  • Codex parity: on Codex a command is realized as its backing skill (Standard sec 3.2). The backing skill’s entry in .codex-plugin/plugin.json is the Codex form. No Codex artifact is generated for the command itself and gen-manifest.mjs is not called for the command.
  • Authoring depth: see the skill’s references/authoring-commands.md.

S7 command-contract errors

Section titled “S7 command-contract errors”

S7 has three distinct failure modes, all errors:

  • Missing description: commands/<name>.md has no description or it is empty. Fix: add a non-empty description to the frontmatter (Standard sec 8.1).
  • Missing maps-to: commands/<name>.md has no maps-to key or it is empty. Fix: add maps-to: <skill-or-workflow-name> to the frontmatter.
  • Unresolved maps-to: the declared maps-to value names no on-disk skill or workflow. Fix: correct the name to match an existing skills/<name>/SKILL.md or _workflows/<name>.md, or create the missing component first.

S7 is conditional: it fires only when commands/ contains at least one command. A plugin with no commands passes S7 automatically.

One-target rule

Section titled “One-target rule”

A command MUST map to exactly one skill or workflow. There is no multi-target maps-to; if a caller needs to route to different backing components, use separate commands.