paradigmxyz/centaur / .agents/skills/prompt-iteration
prompt-iteration
VerifiedTry edited Centaur persona, prompt overlay, or base system prompt changes live from inside a sandbox by running a fresh child agent on the recomposed prompt. Use when iterating on tools/personas/*/PROMPT.md, services/sandbox/SYSTEM_PROMPT.md, or an overlay repo's prompt without redeploying.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.6 KB |
| scripts/compose_preview.shexecutable | script | 3.3 KB |
scripts/compose_preview.sh
View on GitHub#!/usr/bin/env bash
# Compose a sandbox system prompt from edited persona/overlay/base files, the
# same way sandbox startup does, into OUT_DIR/AGENTS.md.
set -euo pipefail
usage() {
cat >&2 <<'EOF'
usage: compose_preview.sh [--base FILE] [--persona FILE | --no-persona] [--repo DIR]... OUT_DIR
--base FILE Use FILE as the base prompt instead of this sandbox's base.
--persona FILE Use FILE as the persona prompt (default: this session's persona).
--no-persona Compose without a persona.
--repo DIR A checkout at .../<org>/<repo> whose services/sandbox/SYSTEM_PROMPT.md
replaces ~/github/<org>/<repo> as an overlay. Repeatable.
EOF
exit 2
}
base="" persona="" no_persona=0 repos=()
while [ $# -gt 1 ]; do
case "$1" in
--base) base="$2"; shift 2 ;;
--persona) persona="$2"; shift 2 ;;
--no-persona) no_persona=1; shift ;;
--repo) repos+=("$2"); shift 2 ;;
*) usage ;;
esac
done
[ $# -eq 1 ] || usage
out="$1"
home="$out/home"
mount="$out/github"
rm -rf "$home" "$mount" "$out/AGENTS.md" "$out/.agents" "$out/.claude"
mkdir -p "$home" "$mount"
copy_if_file() { if [ -f "$1" ]; then cp "$1" "$2"; fi; }
if [ -n "$base" ]; then
# Both names: AGENTS_BASE.md is the selected base, and AGENTS.md lets
# compose drop a mounted checkout whose SYSTEM_PROMPT.md equals the base.
cp "$base" "$home/AGENTS_BASE.md"
cp "$base" "$home/AGENTS.md"
else
copy_if_file "$HOME/AGENTS_BASE.md" "$home/AGENTS_BASE.md"
copy_if_file "$HOME/AGENTS.md" "$home/AGENTS.md"
fi
copy_if_file "$HOME/AGENTS_OVERLAY.md" "$home/AGENTS_OVERLAY.md"
if [ -n "$persona" ]; then
cp "$persona" "$home/AGENTS_PERSONA.md"
elif [ "$no_persona" = 0 ]; then
copy_if_file "$HOME/AGENTS_PERSONA.md" "$home/AGENTS_PERSONA.md"
fi
link_repo() {
local src org_dir
src="$(cd "$1" && pwd)"
# With --base, a checkout still carrying the baked base is the old base,
# not an overlay; compose only dedupes against the new one.
if [ -n "$base" ] && [ -f "$HOME/AGENTS.md" ] \
&& cmp -s "$src/services/sandbox/SYSTEM_PROMPT.md" "$HOME/AGENTS.md"; then
return
fi
org_dir="$mount/$(basename "$(dirname "$src")")"
mkdir -p "$org_dir"
ln -sfn "$src" "$org_dir/$(basename "$src")"
}
for repo in "$HOME"/github/*/*; do
if [ -d "$repo" ]; then link_repo "$repo"; fi
done
for repo in ${repos[@]+"${repos[@]}"}; do
link_repo "$repo"
done
# Expose the session's installed skills, found the same way as
# install_tool_shims._workspace_dir(), to children started in OUT_DIR.
workspace="${WORKSPACE_DIR:-${CENTAUR_WORKSPACE_DIR:-}}"
if [ -z "$workspace" ]; then
workspace="${CENTAUR_STATE_DIR:-$HOME/state}/workspace"
if [ "${CENTAUR_PERSISTENT_STATE:-}" != 1 ] && [ ! -e "$workspace" ]; then
workspace="$HOME/workspace"
fi
fi
if [ -d "$workspace/.agents/skills" ]; then
mkdir -p "$out/.agents" "$out/.claude"
ln -sfn "$workspace/.agents/skills" "$out/.agents/skills"
ln -sfn "$workspace/.agents/skills" "$out/.claude/skills"
else
echo "warning: no skills at $workspace/.agents/skills; the child will lack session skills" >&2
fi
compose-system-prompt --home-dir "$home" --repo-mount "$mount" --target-prompt "$out/AGENTS.md"
[ -f "$out/AGENTS.md" ] || { echo "no base prompt found" >&2; exit 1; }
echo "$out/AGENTS.md"
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |