The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines a framework for processing external data during investigation, which constitutes a surface for indirect prompt injection.
Ingestion points: The methodology requires the agent to examine source code, test files, and git history as defined in the investigation scope within references/hypothesis-testing.md.
Boundary markers: The instructions lack specific delimiters or isolation procedures (such as XML tags or boundary markers) to separate untrusted data from the agent's internal instructions.
Capability inventory: The agent is tasked with tracing causal chains and interpreting the logic of external files, which involves deep processing of potentially untrusted data.
Sanitization: No sanitization, validation, or filtering mechanisms are defined for the content of the files being debugged.
[NO_CODE]: The skill consists exclusively of markdown documentation and templates. No scripts, executables, or configuration files that execute commands were detected.

parallel-debugging