con un clic
ccc-defaults
CCC workspace default behaviors — always applied in this workspace
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Menú
CCC workspace default behaviors — always applied in this workspace
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
CCC HEARTBEAT execution — spawns a background subagent for periodic liveness check and Telegram notification
CCC workspace boot sequence — persona load + background agent orchestration
Manage scheduled jobs — add, remove, pause, resume, edit, list. Live-syncs JOBS.yaml with CronCreate/CronDelete.
Standard flow for receiving tasks from channels (Telegram/Discord etc). Immediate acknowledgment → background execution → progress reporting → completion notification. Always follow this skill when receiving channel messages.
Interactive SOUL.md configuration — persona, identity, language, and personality setup via Telegram
Fetch an OpenClaw/ClawHub skill and convert it to a Claude Code project skill in .claude/skills/
| name | ccc-defaults |
| description | CCC workspace default behaviors — always applied in this workspace |
These rules are always applied in the CCC workspace. They override general defaults.
Always use curl via Bash, never WebFetch.
curl -s "https://api.example.com/endpoint"
Why: WebFetch has a 15-minute cache, making it unsuitable for real-time data (prices, status, etc.). curl always fetches fresh data.
WebFetch is only acceptable for: parsing static documentation pages where caching is harmless.
Every workspace must have git initialized and a .gitignore.
Git init and .gitignore creation are handled by the installer (install.sh/install.bat) on first run.
What TO commit:
.claude/settings.json — base permissions and hooks (no secrets).claude/skills/** — all skill definitionsstart.bat, start.sh — launchers*.example.md — public templatesCommit after every meaningful change.
Trigger: after completing a task that modifies files (config, skills, scripts, docs).
git add <specific files> # never git add -A or git add .
git commit -m "short description"
Security rules (conservative):
git add -A or git add ..env, *.key, *.pem, settings.local.json, memory/, files with tokens/passwords--no-verifyCommit message format: one concise line describing what changed and why.
Always wrap bash commands in cron prompts with a lock file check.
Cron triggers can fire 2–3× in the same interval due to scheduler jitter. Without deduplication, this causes duplicate API calls and duplicate Telegram messages.
Pattern:
LOCK=/tmp/ccc-<job-id>; NOW=$(date +%s); if [ -f "$LOCK" ] && [ $((NOW - $(cat "$LOCK"))) -lt <threshold> ]; then echo "SKIP"; else echo $NOW > "$LOCK"; <your command>; fi
Threshold = ~75% of the cron interval in seconds:
| Interval | Threshold |
|---|---|
*/2 * * * * (2 min) | 90s |
*/5 * * * * (5 min) | 225s |
*/10 * * * * (10 min) | 450s |
0 * * * * (hourly) | 2700s |
In the cron prompt, add: "If output is SKIP, stop here and do nothing."
Why not just ignore duplicates manually? Manual dedup (skipping extra triggers in context) still executes the bash command and hits the API. The lock file prevents the API call itself.