The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses eval and source commands to execute the output of local binaries (gstack-slug, gstack-repo-mode) during its preamble phase. This runtime assembly and execution of code is a medium-risk pattern.
[INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection as it ingests untrusted data from various configuration files (fly.toml, render.yaml, netlify.toml, package.json) and CLAUDE.md. It possesses significant capabilities, including Bash execution and file writing, without implementing explicit boundary markers or sanitization for the ingested content.
Ingestion points: fly.toml, render.yaml, netlify.toml, package.json, .github/workflows/*.yml, and CLAUDE.md.
Boundary markers: Absent.
Capability inventory: Bash, Write, Edit, Read tools.
Sanitization: Absent.
[DATA_EXFILTRATION]: The skill manages a synchronization feature for project "artifacts" (such as plans and reports) to a remote GitHub repository. Additionally, it performs network operations via curl to verify the status of production health check endpoints.
[COMMAND_EXECUTION]: The skill relies heavily on the Bash tool to perform complex environment discovery, manage session state in ~/.gstack/, and interact with the local filesystem and Git repository.

setup-deploy