The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read 'agent instructions' from the target repository as the first step of the procedure. This introduces a surface where malicious instructions in the analyzed project could attempt to hijack the agent's behavior during the installation process.
Ingestion points: Reads local agent instructions in the target repository.
Boundary markers: None specified; instructions are read directly into context.
Capability inventory: The skill has the capability to write to the filesystem (.pre-commit-config.yaml) and execute shell commands (git, pre-commit).
Sanitization: No sanitization or filtering of the ingested instructions is mentioned.
[COMMAND_EXECUTION]: The skill executes several shell commands to manage the installation and validation of the pre-commit hook.
Evidence: Uses git ls-remote to resolve tags and pre-commit run to validate the hook configuration.
[EXTERNAL_DOWNLOADS]: The skill downloads and integrates a pre-commit hook from an external repository.
Evidence: Configures the agent to pull from https://github.com/ruarfff/anti-slop-python.git. This repository belongs to the skill author.

install-anti-slop-python