The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes git to calculate diffs and determine the scope of changes. It also uses node to execute a local script (shipguard.mjs) to manage visual review servers.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from git diff and git status to automatically determine which auditing lanes are applicable. This creates an attack surface where malicious content in the diff could attempt to influence the agent's orchestration logic. Ingestion points: Phase 0 in SKILL.md (via git diff and git status). Boundary markers: The instructions do not specify any delimiters or warnings to ignore instructions within the diff content. Capability inventory: Shell command execution (git, node), file system modifications (visual-tests/_results/), and the ability to trigger other security-sensitive skills (sg-code-audit, etc.). Sanitization: There is no evidence of sanitization or filtering of the diff output before processing.

sg-ship