The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an extreme volume of redundant content across eight reference files (e.g., testing-strategies.md, architecture-patterns.md, state-management.md). Each file contains 150 sections consisting of the same technical boilerplate. This context-stuffing technique serves no functional purpose and is typically used to fill an agent's context window or obscure other instructions.
[PROMPT_INJECTION]: The skill processes untrusted user data via bug reports and root cause analysis documents using implementation patterns defined in SKILL.md.
Ingestion points: User-supplied titles and descriptions are parsed by the BugReport and RCADocument classes in SKILL.md.
Boundary markers: No delimiters or warnings to ignore embedded instructions are present in the parsing logic.
Capability inventory: The skill has access to powerful tools and commands including git bisect, kubectl logs, npm run build, npx jest, and various debuggers (gdb, jstack).
Sanitization: There is no filtering or validation of the input text before it is interpolated into the agent's workflow.
[COMMAND_EXECUTION]: Instructions in references/remote-debugging.md direct the user to use sudo to modify sensitive kernel parameters (core_pattern) and manage core dump generation at a system level.
[COMMAND_EXECUTION]: The skill provides automated scripts in SKILL.md that execute local shell commands, including git bisect run, npm run build, and npx jest, which are triggered during the debugging workflow.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading and installing various third-party debugging tools and libraries, such as debugpy via pip, the delve debugger for Go, and multiple OpenTelemetry observability packages via npm.

dev-loop-debugging-strategy