The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands (e.g., TEST_CMD, LINT_CMD) retrieved from the .release.yml configuration file using eval. This allows project-level configuration to define executable actions performed by the agent within the user's environment.
[COMMAND_EXECUTION]: The skill automatically performs git add -A followed by a commit when uncommitted changes are detected during the pre-flight checklist. This broad staging behavior could lead to the accidental inclusion of sensitive files, local environment variables, or credentials in the repository history if they are not explicitly defined in .gitignore.
[EXTERNAL_DOWNLOADS]: The skill references and suggests the installation of external tools and extensions, specifically sem-cli via Homebrew and the gh-signoff extension from Basecamp's official repository.
[PROMPT_INJECTION]: The skill processes untrusted data from git log and .release.yml to generate changelogs and determine versioning logic, creating a surface for indirect prompt injection.
Ingestion points: Commit messages retrieved via git log and configuration values from .release.yml.
Boundary markers: Absent; commit messages are interpolated directly into changelog templates without specific delimiters or instructions to ignore embedded instructions.
Capability inventory: The skill possesses capabilities for shell execution (eval), file modification (sed, node), and repository management (git commit, git tag, gh release).
Sanitization: No explicit sanitization or filtering is performed on the ingested commit messages or configuration values before processing.

cali-ops-github-releases