Skip to content
Back to skills

Rch

ASecurity

Offload one build through RCH or diagnose its remote compiler. Use when: remote compilation is selected; report errors without creating a retry controller.

  • 447 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 2, 2026
ai-agentsgoapi

Works with

  • api

Security analysis

A100/100

Pro scans all 8 files and shows the line behind each finding

Scanned September 12, 2026

npx -y skills add boshu2/agentops --skill rch --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Rch?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Rch
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/boshu2-rch-4115fee8/badge)](https://www.skillsdirectory.com/skills/boshu2-rch-4115fee8)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: rch
user-invocable: true
skill_api_version: 1
hexagonal_role: supporting
consumes: []
produces: []
context_rel: []
metadata:
  dependencies: []
  capabilities: [rch]
  effects: [remote_compilation_offload, authorized_remote_daemon_worker_mutation]
  canonical_status: canonical
  disposition: keep_optional_adapter
  tier: execution
description: 'Offload one build through RCH or diagnose its remote compiler. Use when: remote compilation is selected; report errors without creating a retry controller.'
practices:
- pragmatic-programmer
output_contract: remote compilation status and diagnostic evidence
---
# RCH — remote compilation specialist

RCH can offload one explicit compilation command or inspect the remote compiler
path. This skill reports what happened; it does not govern retries or repair.

Staged diagnosis works because the offload pipeline fails in order —
availability, configuration, hook, classification, sync, remote compile,
worker pressure — so the first failing stage localizes the fault and every
later stage is noise until it passes. Remediate in irreversibility order:
read-only probes and config inspection before daemon restarts, restarts before
cleanup, and destructive cleanup or worker mutation only with explicit caller
authority.

Named failure mode — **green-local blindness**: a passing `[RCH] local (...)`
build read as offload success; the local fallback hid that the remote claim
was never proved.

Anti-pattern: re-running the build hoping the fallback reason disappears.
Corrective: read the recorded fallback reason and fix that stage; the pipeline
fails deterministically, not moodily.

## Procedure

1. Capture `rch check`, `rch doctor --json`, worker status, and the relevant
   `[RCH]` summary before mutation.
2. For diagnosis, identify the first failing stage: availability, configuration,
   hook, classification, sync, remote compile, or worker pressure.
3. Run only the caller-authorized command or documented safe diagnostic once.
4. Capture the exact command, worker when known, exit code, local-fallback reason,
   and post-action status.
5. Stop and return the evidence.

`[RCH] local (...)` means the requested remote-offload claim was not proved even
when the local build succeeds. Destructive cleanup, worker deployment, daemon
configuration, and remote mutation require explicit caller authority.

`rch check` exit status adjudicates readiness (0 = ready, nonzero = not
offload-ready). Do not read `rch doctor --json` `success: true` as readiness — a
successful diagnostic report can coexist with a down daemon or unreachable
workers. Adjudicate on `rch check`; use `doctor` for the reasons behind it.

## Output

Return a factual packet with status (`remote`, `local_fallback`, `failed`, or
`not_proven`), commands and exit codes, worker, summary line, and checked/not
checked surfaces. Do not include a next action. `not_proven` here is a runtime
diagnosis status, not an AgentOps verdict; it carries no verdict weight and never
substitutes for a `verdict.v2`.

## References

The SKILL.md authority boundary above governs every reference below. Where a
reference lists a remediation, its read-only diagnostics run autonomously but its
remote, privileged, or irreversible steps (remote/`sudo` mutation, daemon
start/restart/reconfigure, worker or fleet deployment, toolchain sync,
destructive cleanup) still require explicit caller authorization first. A
reference never widens the autonomy the kernel grants.


- [Fail-open reasons](references/FAIL_OPEN.md)
- [Error catalog](references/ERROR_CODES.md)
- [Troubleshooting](references/TROUBLESHOOTING.md)
- [Recovery playbooks](references/RECOVERY_PLAYBOOKS.md)
- [Worker operations](references/WORKERS.md)
- [Configuration](references/CONFIGURATION.md)
- [Machine-readable surfaces](references/MACHINE_INTROSPECTION.md)

Files in this skill

  • SKILL.md3.7 KB
  • references/CONFIGURATION.md4.7 KB
  • references/ERROR_CODES.md9.4 KB
  • references/FAIL_OPEN.md13.7 KB
  • references/MACHINE_INTROSPECTION.md7.3 KB
  • references/RECOVERY_PLAYBOOKS.md14 KB
  • references/TROUBLESHOOTING.md12.2 KB
  • references/WORKERS.md3.4 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.