The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions direct the agent to use local tools such as fd, ast-grep, git grep, and bat to perform lexical and structural searches within the codebase. These operations are intended for context gathering to resolve design dependencies and are restricted to local analysis.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted codebase data to inform its questioning protocol.
Ingestion points: Codebase files are read via lexical search (git grep), structural search (ast-grep), and targeted reading (bat) in the process step titled 'Explore the codebase before asking when possible'.
Boundary markers: No explicit boundary markers or instructions to ignore embedded instructions are specified when the agent interpolates file content into its context.
Capability inventory: The skill uses discovery and reading tools (fd, ast-grep, git grep, bat). No file-write, network, or privilege escalation capabilities are present in the skill.
Sanitization: No sanitization or filtering of the ingested code content is performed.
[SAFE]: No evidence of malicious prompt injection, data exfiltration, obfuscation, or unauthorized persistence mechanisms was found. The skill follows a well-defined protocol for user interaction via a single-select question/answer framework.

grill-me