The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements an iterative retrieval pattern that ingests untrusted data from a codebase into the agent's active context, creating a surface for indirect prompt injection.
Ingestion points: The retrieveFiles function and the multi-phase loop (SKILL.md) ingest content from external files.
Boundary markers: Absent. The skill does not instruct the agent to wrap retrieved content in delimiters or use safety warnings to ignore embedded instructions.
Capability inventory: The skill snippets do not include dangerous capabilities like subprocess execution, file writing, or network operations.
Sanitization: Absent. The pattern does not include steps to sanitize or validate the content of retrieved files before processing.
[SAFE]: No obfuscation, hardcoded credentials, or unauthorized network operations were detected. All external references are to well-known services for documentation purposes.

iterative-retrieval