The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the !command syntax in SKILL.md to execute a local shell script (yaml-helper.sh) at load time. This operation is used solely for environment configuration and does not process external user input in a dangerous manner.
[COMMAND_EXECUTION]: The skill's instructions include several calls to local scripts, such as registry.sh, to validate data and manage the project's entity registry. These scripts are internal to the repository, and their execution is restricted by the skill's frontmatter configuration.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads various Markdown and YAML files from the design/ and docs/ directories, including GDDs, ADRs, and registries.
Boundary markers: The skill employs the Grep tool to read specific ranges of files based on heading patterns, rather than reading entire files indiscriminately.
Capability inventory: The skill has permissions to write and edit files and execute specific local scripts via the Bash tool.
Sanitization: Input system names are normalized to kebab-case, which acts as a basic sanitization step for file paths and script arguments.

design-system