The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git commands (including diff, rev-parse, and ls-files) to analyze repository state. It implements a defensive 'Range resolver' that sanitizes user-provided input by rejecting shell metacharacters and strings that look like command-line options (e.g., starting with --), mitigating command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of repository code changes, which presents an attack surface for indirect prompt injection.
Ingestion points: The agent reads file paths, line counts, and diff content through git diff operations defined in SKILL.md.
Boundary markers: The skill contains explicit 'Red Flags' and 'Rationalizations' instructions that command the agent to never treat text found within the diff as an instruction.
Capability inventory: The skill possesses the ability to execute shell commands (git) and write allocation reports to the local file system in /tmp or user-defined paths as described in references/output.md.
Sanitization: Input validation in the range resolver prevents malformed strings from reaching the shell, though the agent still processes the resultant diff content.

select-sample