The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from the system environment and external sources, which could be exploited to influence agent behavior.
Ingestion points: Crash metadata from coredumpctl (SKILL.md, line 14), system logs from journalctl (SKILL.md, line 25), backtraces from gdb (SKILL.md, line 44), and existing GitHub issue content from gh search and view commands (reporting.md, lines 38, 53).
Boundary markers: Absent. The skill does not provide delimiters to separate untrusted data from the agent's instructions.
Capability inventory: The skill possesses capabilities for network access via the gh tool (reporting.md, line 37) and Arch Linux debuginfod server (SKILL.md, line 44), as well as file system interactions through mktemp and coredumpctl (SKILL.md, line 41).
Sanitization: The skill provides defensive shell quoting instructions (SKILL.md, line 75) to prevent command injection, but no sanitization is applied to the content of the crash data or GitHub responses themselves.
[COMMAND_EXECUTION]: The skill makes use of standard system utilities like coredumpctl, gdb, and the GitHub CLI (gh). It fetches debug symbols from Arch Linux's official debuginfod server (https://debuginfod.archlinux.org). The execution of these commands follows security best practices, such as using mktemp for core files and clearing them using trap handlers to protect sensitive memory data.

diagnose-crash