The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted data by searching and reading source code from external repositories which may contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill uses tools like rg, grep, and sed (as specified in SKILL.md) to pull file content into the agent's context.
Boundary markers: The instructions in SKILL.md promote the use of precise line ranges (e.g., sed -n 120,160p) to constrain the amount of ingested data.
Capability inventory: The skill enables file system reads, repository-wide searches, and git history inspection.
Sanitization: No explicit sanitization or filtering of the searched content is performed before it is added to the context.
[SAFE]: No evidence was found of malicious behaviors such as hardcoded credentials, unauthorized network calls, shell persistence mechanisms, or obfuscated payloads.

code-search