Requirements
- Target platform
- OpenClaw
- Install method
- Manual import
- Extraction
- Extract archive
- Prerequisites
- OpenClaw
- Primary doc
- SKILL.md
Use when: hardening OpenClaw cron/background workers (POSIX shells: bash/sh) against brittle quoting, cwd/env drift, and false pipeline failures (SIGPIPE, pi...
Use when: hardening OpenClaw cron/background workers (POSIX shells: bash/sh) against brittle quoting, cwd/env drift, and false pipeline failures (SIGPIPE, pi...
Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.
I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete.
I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run.
A reliability-first checklist for OpenClaw cron workers and any unattended automation.
This skill is POSIX-focused (bash/sh examples). The principles are portable, but if you're on Windows/PowerShell you'll need equivalent patterns.
Many OpenClaw setups treat emitting exactly NO_REPLY as "silent success" (no human notification). If your runtime does not support NO_REPLY, interpret it as: print nothing on success.
Scripts-first: move logic into a repo script (recommended: tools/<job>.py or tools/<job>.sh). One command in cron: cron should run one short command (no multi-line bash -lc '...'). Deterministic cwd/env: cd to the repo (or have the script do it), and document required env vars. Silent on success: print nothing (or exactly NO_REPLY) when OK; only emit a short alert when broken. Also see: references/cron-agent-contract.md references/pitfalls.md
Cron failures are rarely "logic bugs". In practice they're often: brittle shell quoting (bash -lc '...' nested quotes) command substitution surprises ($(...)) one-liners that hide escaping bugs (python -c "...") cwd/env drift ("works locally, fails in cron") pipelines that fail for the wrong reason (pipefail + head / SIGPIPE) The fix is boring but effective: scripts-first + deterministic execution + silent-on-success.
Even on POSIX, do not hardcode deployment-specific absolute paths tied to one machine. Prefer: repo-relative paths environment variables you document minimal wrappers that cd into the repo
Likely causes: unclosed $(...) from command substitution broken nested quotes in bash -lc ' ... ' Fix pattern: Replace the whole multi-line shell block with a script. Cron calls exactly one short command, for example: python3 tools/<job>.py
Symptom: command exits non-zero even though the output you wanted is fine Fix pattern: avoid pipefail when piping into head or better: do the filtering in a script (read only what you need)
Common causes: wrong working directory missing env vars different PATH Fix pattern: cd into the repo (or have the script do it) keep dependencies explicit and documented
Symptom: ! [rejected] ... (non-fast-forward) when automation pushes to a long-lived PR/feature branch. Conservative fix (no force-push): On rejection, fetch the remote branch, transplant your new local commits onto it (cherry-pick), then retry push once.
Use this near the top of a cron prompt (2 lines, low-noise): Hardening (MUST): follow references/cron-agent-contract.md (scripts-first, deterministic cwd, silent-on-success). Also apply the cron-worker-guardrails skill. If parsing/multi-step logic is needed, write/run a small tools/*.py script.
Identity, auth, scanning, governance, audit, and operational guardrails.
Largest current source with strong distribution and engagement signals.