The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted Jenkins console logs through the jenkins_mcp-get_build_console_output tool to determine failing stages and suggest fixes. An attacker controlling the Jenkins output could attempt to influence agent behavior through these logs. This is a common attack surface for log analysis tools and is mitigated by the agent's internal safety guidelines.
Ingestion points: Jenkins console output retrieved via MCP tools (SKILL.md) or curl fallbacks.
Boundary markers: Absent; there are no specific delimiters defined to separate external log content from agent instructions.
Capability inventory: The skill can trigger or stop builds via MCP tools and run shell commands (SKILL.md).
Sanitization: No explicit sanitization of log content is described.
[COMMAND_EXECUTION]: The skill provides documentation for executing kubectl and curl commands. These are intended for troubleshooting the local lab environment and are within the expected scope of a DevOps debugging tool.
[CREDENTIALS_UNSAFE]: The documentation references locations of secret files (e.g., jenkins/mcpserver/secret.yaml) used within the lab environment and describes authenticating with administrative credentials for API access. These references are for infrastructure troubleshooting purposes and do not involve hardcoded secrets.

debug-jenkins