Execution Verifier
WarnAudited by ClawScan on May 10, 2026.
Overview
This skill is transparent about verifying progress, but its recommended loop can automatically force-run an executor job and its helper can run arbitrary shell commands.
Use this only if you intentionally want an autonomous verifier/executor loop. Before enabling it, replace the sample project path and cron UUID, require approval for executor runs, avoid arbitrary shell strings, and set clear limits and stop conditions for any scheduled jobs.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If used as written, the agent may force-run an executor job that changes project files or performs other work whenever the verifier decides progress is missing.
The skill makes a forced cron/executor run the automatic fallback when no progress is detected, without artifact-backed approval, allowlisting, or scope limits.
Behavior: - Step A: verify current progress - Step B: if no progress, auto-trigger executor ... --execute-cmd "openclaw cron run fc567f18-83fa-426c-8181-71a10f4568b3 --force"
Require explicit user approval before running the executor, replace the hard-coded cron job with a user-selected job, remove force/no-confirm defaults, and scope the command to the intended project.
A mistaken, manipulated, or overly broad command string could execute unintended shell operations under the user's local permissions.
Both verifier and executor commands are caller-supplied strings and are executed through the system shell, so the wrapper can run arbitrary local commands.
ap.add_argument("--verify-cmd", required=True, help="command that outputs verifier JSON")
ap.add_argument("--execute-cmd", required=True, help="executor command to run when no progress")
...
p = subprocess.run(cmd, shell=True, capture_output=True, text=True)Avoid `shell=True`, use structured argument lists or an allowlist of known safe commands, and keep executor commands user-reviewed.
The system could continue launching work in the background on a schedule, causing repeated changes or resource use after the user has moved on.
The recommended pattern creates recurring autonomous jobs, including one that performs work and writes files, with no stated lifetime, stop condition, or human review checkpoint.
Use two jobs: - **Executor job (isolated agentTurn, every 30m):** do real work + write files. - **Verifier job (main systemEvent, every 30m offset +5m):** run closed-loop script above.
Make the cron setup opt-in, document how to pause or remove the jobs, and add limits such as maximum runs, project scope, and approval before file-changing actions.