The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions for an agent to perform bulk changes across many files. Processing large amounts of potentially untrusted file content constitutes an attack surface where malicious instructions could be embedded.
Ingestion points: The skill is designed to operate on "many files at once," as described in both SKILL.md and references/current-guidance.md.
Boundary markers: The guidance mitigates risk by requiring explicit verification steps, such as inspecting git diff and ensuring match counts are non-zero.
Capability inventory: The skill leverages subprocess capabilities via shell commands like git and grep to perform its tasks.
Sanitization: The instructions specifically warn against shell word-splitting and recommend using while read loops and arrays to safely handle file paths.
[COMMAND_EXECUTION]: The skill includes bash code examples and references to command-line tools (git, grep) intended for identifying and verifying file changes. These are standard developer tools used here to facilitate verification and avoid silent failures.

bulk-change-verification