The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze and edit Go source code, module manifests, and workspace files, creating a vulnerability surface where instructions embedded in user-controlled content could influence agent behavior.\n
Ingestion points: Processes .go, go.mod, go.sum, and go.work files as specified in the SKILL.md trigger conditions.\n
Boundary markers: Absent; the instructions do not define delimiters to distinguish code content from agent instructions.\n
Capability inventory: The skill uses tools for file modification (Read, Edit, Write) and scoped shell execution (Bash(go:), Bash(git:)) as listed in SKILL.md frontmatter.\n
Sanitization: Absent; the instructions do not mandate specific filtering or escaping of content found in Go files.\n- [SAFE]: The skill incorporates significant security best practices and defensive coding guidelines.\n
The documentation in references/exec.md provides explicit instructions on preventing command injection by using argument vectors instead of shell strings.\n
The references/logging.md guidelines strictly forbid the logging of secrets, tokens, or personal data.\n
The references/database.md instructions promote the use of safe transaction boundaries and parameterized queries to prevent data leaks.\n
The references/optional.md file provides a generic implementation to prevent nil pointer dereference issues.

go-idiomatic