The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides several pre-defined diagnostic snippets designed to be executed in the environment. These include:
Linux process environment inspection via /proc/<pid>/environ.
A Python script to map network sockets to process file descriptors.
Binary string extraction using grep and strings-like patterns.
File access tracing using strace.
Process descriptor and working directory inspection.
[EXTERNAL_DOWNLOADS]: The skill includes a diagnostic command for fetching source code content directly from the GitHub API (gh api repos/<org>/<repo>/contents/...). This targets a well-known service for legitimate troubleshooting and research purposes.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data such as application logs, process outputs, and source code files.
Ingestion points: External data is processed in SKILL.md through the 'Observation' phase of the diagnostic cycle.
Boundary markers: None defined; the skill relies on the agent's internal reasoning to distinguish data from instructions.
Capability inventory: The skill has access to shell execution, Python execution, file reading, and network access via the GitHub API.
Sanitization: No specific sanitization or validation mechanisms are described for the external content being analyzed.

ultrathink-protocol