feat: bring skill content to parity with archagents (closes #15) by rob-archastro · Pull Request #19 · ArchAstro/archastro

rob-archastro · 2026-04-13T02:42:54Z

What changed

Renames, updates, and adds skills so archastro-cli has the same 10 concepts as archagents with consistent kebab-case naming and up-to-date content.

Renames

agent_authoring → author-agent (kebab-case, verb-first — matches archagents)
agent_deploy → deploy-agent (same)

Updated existing skills

author-agent: verb-first commands (describe configsample, describe scriptdocs), local-config-dir routing section
deploy-agent: verb-first commands (deploy configs), configs routing section
chat: quick reference command table, ergonomics improvements
auth / install: minor phrasing alignment
impersonate: already at parity from earlier PRs, small alignment tweaks

New skills (4)

build-script: AgentScript authoring, syntax reference, common mistakes, deployment
build-skill: SKILL.md authoring, frontmatter format, publishing
build-workflow: workflow node types (http, script, branch, approval, loop), deployment
manage-configs: local config management, archastro init --enable-configs, sync/deploy workflows

Manifest updates

All three descriptions updated to mention the new capabilities (build scripts/skills/workflows, manage configs)
Codex defaultPrompt updated
Version bumped 0.3.1 → 0.3.2

Scope

10 source files producing 21 outputs (8 Claude skills + 3 Claude commands + 10 Codex skills), plus 3 manifest description updates. All content ported from archagents with archagent → archastro naming adaptation.

Risk

Medium. Renames agent_authoring and agent_deploy skill directories — users with the old version cached will need /reload-plugins (the version bump forces cache refresh). All 4 new skills are additive.

Testing

python3 scripts/generate-plugin-content.py --check → OK: 21 file(s) in sync with sources/.
python3 scripts/check_plugin_repo.py → all 5 checks pass
Source file names match archagents exactly (10 for 10)
Zero archagent/archagents references in sources or generated output

🤖 Generated with Claude Code

Rename, update, and add skills so both repos have the same 10 concepts with consistent naming and content. Renames: - agent_authoring → author-agent (kebab-case, verb-first) - agent_deploy → deploy-agent (kebab-case, verb-first) Updated existing skills (ported from archagents): - author-agent: verb-first commands, local-config-dir routing section - deploy-agent: verb-first commands, configs routing section - chat: quick reference command table, ergonomics improvements - impersonate: already at parity from earlier PRs - auth: minor phrasing alignment - install: minor phrasing alignment New skills (4): - build-script: AgentScript authoring, syntax reference, deployment - build-skill: skill authoring, SKILL.md format, publishing - build-workflow: workflow authoring, node types, deployment - manage-configs: local config management, init, sync, deploy All 10 sources produce 21 outputs (8 Claude skills + 3 Claude commands + 10 Codex skills). Version bumped 0.3.1 → 0.3.2. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

PR #19 copied the archagents impersonate source which had the full body outside the conditional blocks. The command output bloated from 29 to 164 lines. Restructured so {{#SKILL}} wraps the full routing, identity, and session content while {{#CLAUDE_COMMAND}} wraps only the thin CLI passthrough. Command output is now byte-identical to the pre-PR-#19 version (verified against f76db09). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>