Ship Loop

Run a chained build→ship→verify→notify pipeline for multi-segment feature work. Use when implementing multiple features in sequence, each as a coding agent t...

This skill appears to be what it claims: a self‑healing build→ship→verify pipeline that runs your coding agent and manipulates your git repo. Before installing or running: 1) Inspect SHIPLOOP.yml in any repository you point it at — it will execute agent_command and preflight/deploy scripts as your user. 2) Don’t run on untrusted repos. Treat SHIPLOOP.yml like a Makefile/CI config. 3) Be mindful of the agent you configure — if you supply an agent CLI that needs API keys, provide them separately and understand their privileges; avoid using example flags like '--permission-mode bypassPermissions' unless you understand the implications. 4) Note Ship Loop writes state (SQLite DB, .shiploop/metrics.json, learnings.yml) and may commit learnings.yml back into the repo — consider whether you want that file tracked. 5) The pre-scan found unicode control characters in SKILL.md; open the files in a hex-aware editor or use a tool to strip/inspect control chars to ensure there is no hidden or obfuscated content. If you want higher assurance, run the code in a disposable environment or review the repository’s Python code (shiploop/) and tests; they appear to implement the documented behavior.

Type: OpenClaw Skill Name: ship-loop Version: 5.0.0 The skill bundle implements an autonomous build and repair pipeline that executes arbitrary shell commands provided in a 'SHIPLOOP.yml' configuration file. While the documentation includes a security notice regarding these risks, the implementation uses high-risk patterns, specifically the use of 'eval' on configuration-derived strings in 'scripts/preflight.sh' and un-sandboxed subprocess execution in 'shiploop/preflight.py' and 'shiploop/agent.py'. These represent classic RCE vulnerabilities that could be exploited if the tool is run against an untrusted repository. No evidence of intentional malice, such as data exfiltration or backdoors, was detected.