The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from git commit messages and project documents to generate changelogs.
Ingestion points: Git log outputs and files located in production/sprints/ and design/gdd/.
Boundary markers: The instructions lack explicit delimitation or warnings to the model to ignore instructions embedded within the commit messages or retrieved documents.
Capability inventory: The skill has access to Write for modifying docs/CHANGELOG.md and Bash for repository interaction.
Sanitization: There is no evidence of filtering or escaping for content retrieved from external sources before it is processed by the model.
[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes shell execution at load time via the ! command syntax to retrieve repository information for context.
Evidence: Found !git log --oneline -30 2>/dev/null and !git tag --list --sort=-v:refname 2>/dev/null | head -5 in the YAML frontmatter.
These commands are restricted to gathering status information for the developer workflow and do not process user-supplied arguments at the context-loading stage.

changelog