Core

Use when the user runs /galdr:core to install, uninstall, or check galdr's core rules block on this machine for Claude, Codex, or Antigravity. Run install once per agent to write the marked block into that agent's rules file (or the Claude enabled flag); uninstall to remove it; status to report state without writing.From its SKILL.md

Install

npx -y skills add nyelonong/galdr --skill core

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

11 stars11 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.8 KB, as published. Nobody here has run it

core

Install, remove, or check the galdr core rules block on this machine, per agent.

The core block is the shared §0–§6 bootstrap (memory-first, Iron Laws, routing, voice, precedence). It ships in this plugin at carriers/<agent>.md wrapped between  and  markers. The installer writes that marked block into the agent's rules file; uninstall removes the marked block and preserves everything else.

Agents and targets

Codex → ~/.codex/AGENTS.md (block from carriers/codex.md).
Antigravity → ~/.gemini/AGENTS.md (block from carriers/antigravity.md).
Claude → ~/.claude/galdr/enabled flag file (no block — touch on install, rm on uninstall). The session-start hook reads this flag.

All target paths are overridable by env vars (GALDR_CODEX_TARGET, GALDR_ANTIGRAVITY_TARGET, GALDR_CLAUDE_FLAG, GALDR_CARRIERS_DIR) so the test suite can sandbox them and the real global files are never touched.

The installer script

scripts/core-install.sh is the write mechanism. Run it as: bash scripts/core-install.sh {install|uninstall|status} [agent] (agent is optional. When it's omitted, the invoking agent already knows its own runtime from its own operating context — whether it is Claude Code, Codex, or Antigravity is inherent to what's running it — and should determine that and pass it as [agent] itself. Only ask the user when genuinely unable to tell.)

It writes atomically (temp file then mv). For Codex / Antigravity:

target absent → create the file with the marked block;
markers present → replace the marked region with the block (idempotent — never duplicates, even on repeated install; content outside the marked region is untouched, per Rules below);
target exists without markers → append the marked block, preserving all prior content.

uninstall removes the marked block (markers included) and preserves everything else. status reports per-agent state without writing.

Consent layer (this skill's job — modeled on usage-bridge)

The script writes; this skill owns consent. Before any write, show the user the exact change and ask for explicit confirmation. Do not proceed without it.

For install / uninstall on Codex or Antigravity:

Show the target file path.
Show the exact block to write (or the exact block being removed).
Show a before/after diff of the target file:
- install over prior content → diff shows prior content preserved, block appended;
- install over markers → diff shows the marked block replaced (not duplicated);
- uninstall → diff shows the marked block gone, prior content preserved.
Ask for explicit confirmation. On confirm, run the script.

For install / uninstall on Claude:

Show the flag file path (~/.claude/galdr/enabled).
Show the action: touch (install) or rm -f (uninstall).
Ask for explicit confirmation. On confirm, run the script.

status writes nothing — no consent needed. Run the script and report its output verbatim.

Rules

Never write before the user confirms. The script is fast; consent is the gate.
Never edit the carrier files. They are the source of truth; the script extracts the marked block from them as-is.
Never duplicate the block. If markers are present, replace; only append when markers are absent.
Preserve all non-galdr content. Uninstall removes only the marked block.