agentleFS
UtsavBalar1231UtsavBalar1231/oh-my-claudeagent/skills/omca-setup/SKILL.md

omca-setup

Skill3 starsChanged 32 days agoMITView on GitHub

Use to install, check, repair or uninstall oh-my-claudeagent's setup in ~/.claude/ (dependencies, runtime, status line), or when the user says \"setup omca\".

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No flag: SETUP MODE

    line 32
  2. Distinctive:

    Run each command and compare the version it prints:

    line 38
  3. Very specific:

    When ast-grep is missing, run sg --version instead.

    line 56
  4. Very specific:

    Ask with AskUserQuestion whether that output shows four small icons or boxes and blanks.

    line 74
  5. Distinctive:

    When they see boxes, preview the change:

    line 76
  6. Very specific:

    Ask with AskUserQuestion whether to apply the actions (apply or skip).

    line 149
  7. Very specific:

    Write only through scripts/setup-statusline.ts, after the user confirms the diff it printed, through the Edit in phase 5, and through …

    line 155
  8. Distinctive:

    Never edit project, local or managed settings.

    line 156

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

omca-setup

Checks what OMCA needs, asks whether the terminal font draws Nerd Font icons, then points the status line and the subagent status line at OMCA's renderer. The only file outside the plugin it writes is the user settings file (the two status line keys and env.OMCA_GLYPHS), plus the launcher omca/statusline.ts beside it, and only after the user confirms the printed change. OMCA's server delivers the orchestration guidance on each session's first prompt, so setup writes nothing into CLAUDE.md.

The user settings directory is $CLAUDE_CONFIG_DIR when that variable is set and not empty, and ~/.claude otherwise; the launcher goes under the same directory. In bash the settings file is "${CLAUDE_CONFIG_DIR:-$HOME/.claude}/settings.json". In PowerShell it is "$(if ($env:CLAUDE_CONFIG_DIR) { $env:CLAUDE_CONFIG_DIR } else { "$HOME\.claude" })\settings.json". SETTINGS below stands for that quoted path in the shell the session uses.

Policy baseline: Claude Code's native settings are authoritative. Managed settings are non-overridable policy, and keys such as allowManagedPermissionRulesOnly, allowManagedHooksOnly and sandbox.failIfUnavailable belong there; this skill never writes or enforces them. OMCA's hooks never auto-allow a command, so allow decisions come from the permission rules in the user's settings, and /fewer-permission-prompts proposes them from past transcripts. The mod's guard denies catastrophic removals outright. For another destructive Bash command it asks in a review dialog when the guardMode option is dialog and a dialog can show; otherwise it denies a command that discards work.

Mode detection

Parse $ARGUMENTS:

  • --uninstall: UNINSTALL MODE
  • --migrate: MIGRATE MODE
  • --check or --doctor: tell the user to run /omca doctor, which reports the mod, the client and bun versions, the hooks, ast-grep, the plugin options, effort and model overrides, the advisor and the status line. Stop there.
  • No flag: SETUP MODE

Setup mode

Phase 1: dependency check

Run each command and compare the version it prints:

claude --version

PASS at 2.1.300 or later. Otherwise FAIL and stop: OMCA's mod needs Claude Code 2.1.300 or later.

bun --version

PASS at 1.4.2 or later. Otherwise, including when bun is missing, FAIL and stop: OMCA's server, its hooks and the status line run on bun.

ast-grep --version

When ast-grep is missing, run sg --version instead. Either one is PASS. Neither is WARN, not a stop: only the structural code search tools need it.

Phase 2: runtime check

Call the health_check tool, loading it if needed: ToolSearch({query: "select:mcp__plugin_oh-my-claudeagent_omca__health_check", max_results: 1}). When runtime is ok, report PASS. Otherwise report the runtime value and runtime_reason verbatim, and continue: the status line works without the runtime. When the tool is still missing, OMCA's server is not connected; report that and continue.

Phase 3: glyphs

The /omca pane, the band and the status line draw Nerd Font icons unless OMCA_GLYPHS names unicode or ascii. No surface can see the terminal's font, so ask.

  1. Print the test line with the script, so the user's terminal draws it:

    bun "${CLAUDE_PLUGIN_ROOT}/scripts/setup-statusline.ts" --glyph-test

    It prints four Nerd Font icons: a check in a circle, gears, an eye and a terminal prompt. They are private-use characters that do not survive being retyped, so never write them into your own reply; point the user at the command's output.

  2. Ask with AskUserQuestion whether that output shows four small icons or boxes and blanks. When they see the icons, report nerd and end the phase.
  3. When they see boxes, preview the change:

    bun "${CLAUDE_PLUGIN_ROOT}/scripts/setup-statusline.ts" --settings SETTINGS --glyphs unicode --glyphs-only

    It prints Already configured: ... when the settings already name unicode; report that and end the phase. Otherwise show the printed diff unchanged and ask with AskUserQuestion whether to apply it. On yes, run the same command with --yes; on no, change nothing and print the command. unicode keeps every pane and status line glyph inside common monospace fonts; ascii is for plain-text output and screen readers.

Phase 4: status line

The statuslineMode plugin option is ${user_config.statuslineMode}. Claude Code substitutes the option into this file only when the user has set it, so when that still reads as a placeholder instead of off or on, the default on applies. When it is off, skip this phase.

  1. Preview the change:

    bun "${CLAUDE_PLUGIN_ROOT}/scripts/setup-statusline.ts" --settings SETTINGS

    It writes nothing. It prints Already configured: ... when there is nothing to do; report that and end the phase. Otherwise it prints a unified diff of the settings file and the launcher it would copy. It sets statusLine to the absolute bun path plus the launcher's path with padding: 1, refreshInterval: 5 and hideVimModeIndicator: true, and subagentStatusLine to the same command with --subagent. Any other statusLine or subagentStatusLine is replaced, and every other key and byte of the file stays as it is. When it exits 1, report its one-line reason and end the phase.

  2. Show the user the printed diff unchanged, then ask with AskUserQuestion whether to apply it. Say that the previous file is kept beside it as settings.json.omca-bak.
  3. On yes, apply it:

    bun "${CLAUDE_PLUGIN_ROOT}/scripts/setup-statusline.ts" --settings SETTINGS --yes

    On no, change nothing and print that command so the user can run it later.

The launcher runs the renderer of the enabled plugin install that Claude Code recorded most recently, under any marketplace, so a plugin update needs no second setup run. In a --plugin-dir checkout, with no installed version, the status line reads omca: no installed plugin version found.

Phase 5: force-style opt-out

The disableForceOrchestrationStyle option is ${user_config.disableForceOrchestrationStyle}; a placeholder means the default false. When it is not true, skip this phase.

Only the installed cache copy is edited, because a --plugin-dir checkout is tracked source. When ${CLAUDE_PLUGIN_ROOT} lies under plugins/cache/ in the user settings directory, Read ${CLAUDE_PLUGIN_ROOT}/output-styles/omca-default.md. If its frontmatter has the line force-for-plugin: true, remove that line with Edit and report that the user's own outputStyle now takes precedence. If the line is already gone, report that. If Edit refuses the path because permissions.blockWritesOutsideWorkingDirectories is on, report that the opt-out needs that setting off for this one edit, and end the phase. A plugin update restores the line, so this phase has to run again after each update. Outside the cache, report that the opt-out applies only to an installed copy.

Report

=== oh-my-claudeagent setup ===
Claude Code  PASS 2.1.300
bun          PASS 1.4.2
ast-grep     PASS 0.44.0 | WARN not found
Runtime      PASS | <runtime>: <runtime_reason>
Glyphs       nerd | unicode written | already unicode | declined
Status line  configured | already configured | declined | off | <reason>
Force style  stripped | already stripped | skipped

Uninstall mode

  1. Preview what setup wrote:

    bun "${CLAUDE_PLUGIN_ROOT}/scripts/setup-statusline.ts" --settings SETTINGS --uninstall

    It removes a statusLine or subagentStatusLine only when it runs OMCA's launcher, the launcher file itself, and OMCA_GLYPHS from env. When it prints Nothing to remove: ..., report that and go to step 3.

  2. Show the diff, ask with AskUserQuestion, and on yes run the same command with --yes.
  3. Tell the user that the plugin itself is removed with /plugin uninstall oh-my-claudeagent@omca, and its marketplace with /plugin marketplace remove omca. This skill does not run either, and it leaves .omca/ in each project alone.

Migrate mode

Moves a project's agent memories, and the agent ids in the user settings, to the names the roles carry now. It runs per project: tell the user to run it in each project that has agent memories. It never runs from the doctor or at session start.

  1. Load the tool: ToolSearch({query: "select:mcp__plugin_oh-my-claudeagent_omca__agents_migrate", max_results: 1}). When it is still missing, OMCA's server is not connected; report that and stop.
  2. Call agents_migrate with no apply. It changes nothing and returns actions, collisions and mentions. Show them as a short list: each action, each collision with its reason, and each mention, which is a file that names an agent from the rename table and that the tool leaves for the user to edit. When all three are empty, report that there is nothing to migrate and stop.
  3. Ask with AskUserQuestion whether to apply the actions (apply or skip). On skip, change nothing.
  4. On apply, call agents_migrate with apply: true and report the actions it took. When a MEMORY.md collided, ask with AskUserQuestion whether to merge the indexes, and on yes call it again with apply: true and merge_indexes: true.

Constraints

  • Write only through scripts/setup-statusline.ts, after the user confirms the diff it printed, through the Edit in phase 5, and through agents_migrate with apply: true, after the user confirms the listed actions.
  • Never edit project, local or managed settings.
  • A second run changes nothing that is already configured.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.