The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run() to execute shell commands in several scripts.
In scripts/monitor.py, it calls an external search tool (python3 scripts/search.py) passing user-configured queries as arguments.
In scripts/setup_cron.py, it executes crontab commands to manage system-level scheduled tasks.
[PERSISTENCE]: The script scripts/setup_cron.py facilitates the creation and modification of user cron jobs. This mechanism ensures that the skill's monitoring and digest scripts run persistently in the background at hourly, daily, or weekly intervals.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content retrieved from external search engines (titles, snippets, and URLs).
Ingestion points: Data enters the system in scripts/monitor.py via the search_topic function, which fetches results from the web-search-plus tool or mock sources.
Boundary markers: The alert message generation in scripts/monitor.py and digest generation in scripts/digest.py lacks clear delimiters or warnings to the AI agent to ignore potentially malicious instructions embedded within the search results.
Capability inventory: The skill possesses capabilities for command execution (subprocess.run), external network communication (requests.post), and local file system modifications (writing to .findings/ and state files).
Sanitization: There is no significant sanitization or escaping of the ingested web content before it is interpolated into summaries and alerts.
[DATA_EXFILTRATION]: In scripts/monitor.py, the skill is designed to send alerts to external Discord webhooks. While intended for user notification, this channel could be used to exfiltrate sensitive search results or data extracted from the system if the configuration is compromised or if malicious search results are processed.

topic-monitor