docs: add CLI agent skills for non-Mogplex hosts#23
Conversation
Adds an Anthropic-spec skill set that teaches external agents (Claude
Code, Agent SDK, Cursor, opencode) how to drive a user's Mogplex
workspace through the CLI as a substitute for a Mogplex MCP server.
- skills/{using-mogplex-cli,mogplex-exec,mogplex-slash,mogplex-auth}/SKILL.md
as canonical source
- content/docs/cli/skills/* mirrors them in the Fumadocs sidebar
- cli/index.mdx gets a Skills card; cli/meta.json includes the section
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![mogplex[bot]](https://avatars.githubusercontent.com/in/3107927?s=60&v=4){kind=small}
There was a problem hiding this comment.
Mogplex PR Review
Status: Attention needed
This is a well-structured documentation-only PR adding Anthropic Agent Skills for the Mogplex CLI. The content is clear, internally consistent, and follows the stated spec. There are no security or correctness issues with the documentation itself. I found three issues worth flagging: a truncated/incomplete .mdx file in the diff, a subtle agent-instruction gap in the preflight check, and minor content drift between the canonical SKILL.md files and their content/docs mirrors.
2 findings were added inline.
Suggestions
- Content drift:
mogplex-authSKILL.md contains a 'Hard rules' section absent from the .mdx mirror (content/docs/cli/skills/mogplex-auth.mdx)
The canonicalskills/mogplex-auth/SKILL.mdincludes a prominent## Hard rulessection (never read auth.json, never echo a raw key, etc.) that does not appear incontent/docs/cli/skills/mogplex-auth.mdx. The PR states the docs pages "mirror" the canonical skill files, but this section is user-safety-relevant and worth surfacing in the rendered docs too. Either add the section to the.mdxor explicitly note in the Callout that the.mdxis an abridged version. The same pattern — SKILL.md 'See also' linking to an authentication guide (https://mogplex.dev/cli/guides/authentication) — is also absent from the mirror page.
|
|
||
| - [mogplex-exec](/cli/skills/mogplex-exec) | ||
| - [mogplex-auth](/cli/skills/mogplex-auth) | ||
| - [Slash commands guide](/cli/guides/slash-commands) |
There was a problem hiding this comment.
Warning: mogplex-slash.mdx is truncated mid-sentence in the diff
The patch for mogplex-slash.mdx ends at line 108 with - Running — the sentence is cut off and the closing content is missing from the diff hunk. This likely means the file itself is complete on the branch but GitHub's diff is clipping it. Please verify the file ends cleanly (e.g. a complete ## See also section) and that the rendered page does not show a fragment. Similarly, using-mogplex-cli.mdx in the diff ends with - The agent cannot read ~/.mogplex/auth.json. All credential state must — also mid-sentence in the patch. Worth an explicit sanity check against the full files on the head branch.
| Before any command that talks to Mogplex, verify the CLI is installed and authenticated: | ||
|
|
||
| ```bash | ||
| command -v mogplex >/dev/null || echo "mogplex CLI not installed" |
There was a problem hiding this comment.
Warning: Preflight command -v mogplex silently continues on failure
The preflight snippet is:
command -v mogplex >/dev/null || echo "mogplex CLI not installed"
mogplex login statusIf mogplex is not on PATH, this prints a warning and then immediately runs mogplex login status anyway, which will produce a confusing command not found error rather than a clean stop. The skill instructs the agent to check for the CLI, but the shell idiom doesn't actually gate the subsequent command. Consider revising to something like:
command -v mogplex >/dev/null 2>&1 || { echo "mogplex CLI not installed — install it first"; exit 1; }
mogplex login statusor make the intent explicit in prose: "If the first command prints the warning, stop here and tell the user to install the CLI before continuing." This is an agent-instruction issue — agents following the snippet literally may try to run mogplex login status even when mogplex is absent. The same snippet is mirrored in content/docs/cli/skills/using-mogplex-cli.mdx and should be updated there too.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: aefea6c679
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| ```bash | ||
| # Global install | ||
| cp -R skills/* ~/.claude/skills/ |
There was a problem hiding this comment.
The global install command assumes ~/.claude/skills/ already exists, but for first-time users that directory is often missing, so cp -R skills/* ~/.claude/skills/ fails with "No such file or directory" and none of the skills are installed. Add a mkdir -p ~/.claude/skills step (or use an install command that creates parents) so the documented install flow works on a clean machine.
Useful? React with 👍 / 👎.
1 participant
Summary
skills/*/SKILL.mdthat teaches external agents (Claude Code, Agent SDK, Cursor, opencode) how to drive a Mogplex workspace through themogplexCLI — a drop-in substitute for a Mogplex MCP server, without asking users to paste tokens into JSON configs.content/docs/cli/skills/), with a new card on the CLI landing page.name+description,namematches directory, bodies start with a single H1, descriptions use the canonical[what]. Use when [trigger]pattern (261–367 chars, well under the 1024 cap).Skills
using-mogplex-climogplex-exec--json/--jsonloutput.mogplex-slashmogplex slash list --json) and run (mogplex exec "/...") slash commands.mogplex-authTest plan
pnpm lintpnpm types:checkpnpm build(all 165 pages generated, includes new/cli/skills/*routes)/cli/skills/mogplex-auth) for callout + code block rendering