agentleFS
zeikarzeikar/hyperclaude/skills/hyper-setup/SKILL.md

hyper-setup

Skill5 starsChanged 48 days agoMITView on GitHub

Diagnose hyperclaude prerequisites and report fixes

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 3 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run the doctor probe. Parse the JSON it emits and report per-check status plus the overall verdict.

    line 10
  2. Very specific:

    If ok is true (no hard FAIL): All hard prerequisites met.

    line 39
  3. Very specific:

    If ok is false (one or more hard FAILs): N hard prerequisite(s) failing — hyperclaude will not work until fixed.

    line 40
  4. Very specific:

    Do NOT spawn Codex, the bridge (codex-bridge.mjs), or any agent.

    line 47
  5. Distinctive:

    Do NOT auto-install missing tools or modify env variables.

    line 48
  6. Distinctive:

    Do NOT treat the Claude Code version WARN as a hard failure.

    line 50
  7. Very specific:

    No manifest entry needed — Claude Code auto-discovers skills/*/SKILL.md.

    line 52
  8. Very specific:

    Do not widen allowed-tools beyond Bash(node:*), Read.

    line 53

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

hyper-setup

Run the doctor probe. Parse the JSON it emits and report per-check status plus the overall verdict.

Prerequisite probe: !node "${CLAUDE_PLUGIN_ROOT}/scripts/setup-doctor.mjs"

What it checks

The probe emits one JSON line: { ok, checks: [{name, detected, required, status, severity, remediation}] }.

The five checks:

  1. Node.js >= 18 — severity: hard. hyperclaude's bridge is stdlib Node; versions below 18 are unsupported.
  2. codex-cli >= 0.130.0 on PATH — severity: hard. Version-floor check only (no capability probe). The bridge spawns codex exec; the tool must be present and at a known-good version.
  3. git on PATH — severity: hard. The bridge reads git state for slug generation and diff targets.
  4. codex --search global flag (pre-subcommand) — severity: hard. The bridge passes --search as a global flag before the subcommand on every Codex spawn; codex-cli must accept codex --search exec --help (exit 0).
  5. Claude Code >= 2.1.232 — severity: conditional. 2.1.232 is the known-good floor for the background-agent transport hyper-plan-loop, hyper-implement-loop, hyper-docs-loop, and hyper-auto (which chains hyper-plan-loop → hyper-implement-loop) use to reach their agent. Surfaced, not gated: a below-floor or unknown version (claude off PATH, unparseable output) is a WARN that never flips the verdict and blocks nothing — the full research→plan→implement flow works regardless.

Reporting directive

The inline probe above runs automatically; parse the single JSON line it emitted to stdout (do not run it again).

The JSON's shape picks the report. A checks[] array — including the normal ok:false "prerequisites failing" case — gets steps 1–3. Only a probe that itself failed (no parseable JSON, or an {ok:false,error:...} object with no checks[]) gets the step 4 fallback.

For the checks[] path:

  1. Present a per-check table with columns: Check, Detected, Required, Status. One row per entry in checks[].
  2. For every check where status is not PASS, print the check's remediation field verbatim on its own line, prefixed with Fix:.
  3. Conclude with an overall verdict line:
    • If ok is true (no hard FAIL): All hard prerequisites met.
    • If ok is false (one or more hard FAILs): N hard prerequisite(s) failing — hyperclaude will not work until fixed. (N = count of checks[] entries with severity:"hard" and status:"FAIL").
    • The Claude Code version WARN must never flip the overall verdict to fail.
  4. Error fallback — print exactly this, and nothing that reads as a pass: Prerequisite probe could not complete: <error or "no parseable output">. hyperclaude prerequisites are UNKNOWN — re-run /hyperclaude:hyper-setup or run the doctor script directly.

Anti-patterns

  • Do NOT spawn Codex, the bridge (codex-bridge.mjs), or any agent. This skill only runs the doctor probe.
  • Do NOT auto-install missing tools or modify env variables. This skill is read-only; report-and-advise only.
  • The Bash(node:*) filter permits arbitrary node scripts, so never edit this skill to invoke anything beyond the doctor probe — read-only is prompt-enforced, not tool-enforced.
  • Do NOT treat the Claude Code version WARN as a hard failure.
  • No npm dependencies.
  • No manifest entry needed — Claude Code auto-discovers skills/*/SKILL.md. disable-model-invocation: true keeps this skill explicit-invoke-only (/hyperclaude:hyper-setup); it is never auto-triggered by its description.
  • Do not widen allowed-tools beyond Bash(node:*), Read.

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.