The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze Go source code (files matching **/*.go), which represents a source of untrusted data. This creates an attack surface for indirect prompt injection where malicious instructions embedded in comments or strings within the audited code could attempt to override the agent's behavior or manipulate its audit findings.
Ingestion points: The agent reads and processes Go source files matching the pattern **/*.go as defined in the SKILL.md frontmatter.
Boundary markers: The instructions do not explicitly define boundary markers or include specific 'ignore embedded instructions' directives when processing the audited source code.
Capability inventory: The skill allows access to powerful tools including Bash(go:*), Write, Edit, and Agent (allowing the spawning of sub-agents).
Sanitization: There is no evidence of sanitization or filtering of instructions found within the target Go source code during the auditing process.
[EXTERNAL_DOWNLOADS]: The skill performs installations of security tools from official and well-known sources during setup.
Evidence: The install field in SKILL.md triggers a go install of golang.org/x/vuln/cmd/govulncheck@latest. Additionally, the 'Tooling & Verification' section suggests installing github.com/securego/gosec/v2/cmd/gosec@latest. Both sources represent trusted development organizations and well-known security utilities.

golang-security