The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) as it processes log files that can contain untrusted data influenced by external attackers (e.g., malicious payloads in user-agent strings or request data that trigger logged errors).
Ingestion points: The skill reads *.log files from the filesystem (SKILL.md).
Boundary markers: No specific boundary markers or instructions are provided to help the agent distinguish between log data and instructions.
Capability inventory: The agent has the capability to write files and execute terminal commands, specifically the gh (GitHub CLI) for creating issues, labels, and projects.
Sanitization: There is no evidence of sanitization or escaping for the log content before it is processed or interpolated into issue templates.
[DATA_EXFILTRATION]: The skill explicitly extracts sensitive information including 'User IDs, request data, and SQL queries' from production logs and includes them in GitHub issues. While this is the intended functionality, it poses a significant risk of exposing production secrets or PII if the GitHub repository is public or shared with unauthorized individuals.
[COMMAND_EXECUTION]: The skill uses the gh (GitHub CLI) to automate administrative tasks such as creating labels (gh label create) and managing project boards (gh project create). Automated modification of repository settings based on potentially attacker-controlled log data increases the risk profile of the skill.

log-monitor