The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various local scripts and command-line utilities including node, git, gh (GitHub CLI), sbt, and a custom shell script ./scripts/build.sh. These operations are limited to the repository context and are necessary for the declared purpose of building and preparing a software release.
[PROMPT_INJECTION]: The workflow involves processing untrusted data from external sources, specifically GitHub release notes via the gh api and the project's CHANGELOG.md file.
Ingestion points: Content is pulled from gh api repos/<owner>/<repo>/releases/generate-notes and the local ./CHANGELOG.md file (SKILL.md).
Boundary markers: The instructions do not define explicit boundary markers or delimiters for the ingested content.
Capability inventory: The agent has capabilities to execute shell commands (node, git, gh) and build scripts (build.sh).
Sanitization: There is no mention of sanitizing or escaping the external content before processing.
Mitigation: The skill significantly reduces risk by requiring explicit maintainer approval for any side-effecting commands such as git push or gh release create.

cosmo-release