The Agent Skills Directory

[PERSISTENCE]: The skill is designed to create and install crontab entries, allowing for automated, recurring execution of code on the host system. While it includes a requirement for user confirmation before installation, this capability facilitates system-level persistence.
[REMOTE_CODE_EXECUTION]: The skill relies on an external script located at ~/.claude/scripts/crontab-manager.py (or alternatively in a local scripts/ directory) to manage crontab mutations. This script is not included within the skill package, making its contents unverifiable and its execution a dependency on external code.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to read and audit shell scripts from the host repository. This creates a surface for indirect prompt injection if an attacker-controlled script contains instructions designed to manipulate the agent's behavior during the audit process.
Ingestion points: Shell scripts discovered in scripts/*.sh, cron/*.sh, jobs/*.sh, and bin/*.sh, as well as any files they source (found in SKILL.md Phase 2).
Boundary markers: None explicitly instructed for the script reading process; the agent is instructed to read the entire file content.
Capability inventory: Uses Bash tool to perform searches and executes the external crontab-manager.py script to modify system state.
Sanitization: The skill attempts to use regex patterns to identify best practices and notes if matches occur inside comments, but there is no formal sanitization of untrusted script content before it is processed by the agent.

cron-automation