The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/tag_scan.py uses the subprocess module to execute the chromium browser and the passe utility.
[COMMAND_EXECUTION]: The skill provides instructions in SKILL.md and checks in scripts/tag_scan.py that prompt the user or agent to use sudo to write a configuration file to /etc/chromium/policies/managed/doh-escape-hatch.json. This is intended to enable DNS-over-HTTPS but involves administrative access to the system.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it analyzes content from arbitrary external websites and processes network data that could contain malicious instructions.
Ingestion points: Network request hostnames and URLs are ingested from the target website by scripts/tag_scan.py and classified.
Boundary markers: Absent; there are no specific delimiters or instructions to ignore embedded commands within the analyzed network data.
Capability inventory: The script can execute shell commands via subprocess and provides instructions for elevated command execution via sudo.
Sanitization: Network data is extracted via urlsplit but is not sanitized or escaped before being processed for reporting or classification.

tamis