The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code and project files to generate documentation, creating a potential surface for indirect prompt injection attacks.
Ingestion points: The skill ingests project source code, README, CONTRIBUTING, and CHANGELOG files to determine project type and generate content (SKILL.md Step 1, 2, and 3).
Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions when reading external code files.
Capability inventory: The agent is empowered to write documentation comments directly to Go source files and update project files such as README.md and llms.txt.
Sanitization: No explicit sanitization or filtering is performed on data extracted from the code before it is interpolated into documentation.
[COMMAND_EXECUTION]: The skill provides documentation templates that include commands for installing standard Go documentation tools.
Evidence: The templates include commands for installing tools from well-known repositories, such as go get -tool github.com/swaggo/swag/cmd/swag@latest (references/application.md) and go tool pkgsite (references/library.md).
[EXTERNAL_DOWNLOADS]: The skill provides templates for downloading project binaries from GitHub's release infrastructure.
Evidence: The delivery documentation section suggests providing users with a curl command to download binaries from official GitHub release paths (SKILL.md Step 10).

golang-documentation