The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the exclamation-backtick syntax (!date, !go version) to capture system information when the skill is loaded. These commands are used legitimately to provide a timestamp and environment context for the audit report.
[INDIRECT_PROMPT_INJECTION]: The skill processes external manifest files and tool outputs, which constitutes an indirect prompt injection surface.
Ingestion points: go.mod, package.json, and the output of security scanning tools like npm audit and govulncheck (SKILL.md).
Boundary markers: None present in the instructions to separate untrusted data from agent instructions.
Capability inventory: Execution of shell commands for auditing and the ability to write to the file system via the obsidian_append_content tool (SKILL.md).
Sanitization: No explicit sanitization of package names or version strings from manifest files is performed before processing.
[UNVERIFIABLE_DEPENDENCIES]: The skill utilizes npx license-checker to perform license audits. This command fetches the license-checker package from the official NPM registry if not already present. This is a well-known tool used for its intended purpose.

deps