H2cv — Development skill for Claude Code
A command-line tool that lets Claude Code start Claude Code sessions and give them instructions.
How to install H2cv
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open sakamossan/h2cv and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What H2cv does
A command-line tool that lets Claude Code start Claude Code sessions and give them instructions.
Alternatives in Development
- Claude Subconscious — Give Claude Code a subconscious 2.9k ★
- Claude Supermemory — Persistent memory across sessions and projects using Supermemory 2.3k ★
- Clawk — Give coding agents a disposable Linux VM, not your laptop 996 ★
README
h2cv (herdr sends to claude verifiably)
A command-line tool that lets Claude Code start Claude Code sessions and give them instructions.
It drives herdr underneath, and closes, in deterministic code, the problems you are bound to hit when you work the claude input box through it.
Why
herdr ships a skill file for AI agents, but a skill file on its own will not let one claude give instructions to another claude. Your agent may well get a message delivered on the fourth attempt — by which point its context is packed with pane ids and screen dumps, and it has lost the thread of whatever it was working on in the first place. The claude input box is a more complicated place than it looks. A short list of what is waiting there:
The completion menu eats your Enter
Claude Code A sends the `/review` skill to Claude Code B with `pane run`. B does not necessarily start working. A wonders why nothing is happening and goes back to look at the pane — again, and again. The Enter never reached the prompt: the completion menu opened on the leading `/` and swallowed it, and the body is still sitting in the input box. Getting out of this takes a procedure rather than a retry, and the obvious move — send the whole thing again — is the wrong one.
`idle` lies for about a second
Right after startup, herdr reports the agent as `idle`. Trust that, fire `agent send` immediately, and the instruction vanishes without ever running: the TUI is still initializing, and everything typed into that window is discarded. The status is not so much wrong as answering a different question than the one you asked, which is why "check the status, then send" is itself the trap.
"It didn't arrive" has three different endings
Nothing arrived, the body arrived without a submit, or the submit landed late. The agent believes it is looking at one condition and re-fires.
Related Skills
Telepath
Give your Claude Code and Codex sessions telepathy — a zero-dependency message bus that lets local coding agen
SafeClaw
to make running containerized Claude Code sessions easy. It lets you spin up multiple isolated sessions, each
Embodify
Give your agent a body: Embodify lets it control robots directly and become your embodied agent. Your best emb
Claude Engineer
A version of command line interface(CLI) that leverages the power of Claude Opus(3-4.7). This framework lets C
Botbus Protocol
The BotBus protocol: how a menu bar app on your computer and the companion apps on your phone and watch stay i
Lets Start Skill
A Claude Code Skill to Make Building Even Easier
Related Agents
Aiskills Architect
Shapes features and systems before a line of implementation is written — names the problem first and lets it p
Part 13: Memory Bridge (Give Coding Agents Your Brain)
When you spawn Codex or Claude Code to build something, they start blind. They don't know your architecture de
Release Notary
Give a compact release-readiness view grounded in doctor findings, sessions, and promotion evidence
