The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external project configuration files to infer project names and slugs.
Ingestion points: Processes package.json, pyproject.toml, Cargo.toml, and go.mod files from the project directory.
Boundary markers: No explicit instruction delimiters are used for the interpolated data.
Capability inventory: The skill writes configuration files to the .devcontainer/ directory and specifies commands for execution in postCreateCommand.
Sanitization: The skill applies slugification (lowercase conversion and character replacement) when generating names for persistent volumes, which provides basic sanitization against malformed input.
[DYNAMIC_EXECUTION]: The skill generates shell scripts (install.sh) and Python scripts (post_install.py) that are subsequently executed during the container initialization process. These scripts are generated based on identified language stacks and project needs.
[COMMAND_EXECUTION]: Requests NET_ADMIN capabilities within the container environment to support network isolation tools like iptables and ipset. This is documented as part of the container's specialized networking configuration.

devcontainer-setup