The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Git repositories and external ticket trackers, creating a surface for indirect prompt injection attacks.
Ingestion points: In SKILL.md, the agent is instructed to read branch names, commit messages, diffstats, and external ticket tracker information (titles, status).
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when reading or reporting this data.
Capability inventory: The skill in SKILL.md utilizes capabilities for git branch, git commit, git push, and git worktree management.
Sanitization: While SKILL.md advises against staging credentials, it lacks specific sanitization or validation steps for textual content retrieved from external sources.

salvage