The Agent Skills Directory

[SAFE]: The skill incorporates a proactive security scanning step in references/safety-protocols.md that uses grep to detect secrets like API keys, AWS credentials, and private keys in staged changes before they can be committed.
[COMMAND_EXECUTION]: Executes shell commands to automate repository management. Several workflows, such as in references/workflow-pr.md, use unquoted shell variables to handle branch names and PR titles. This practice is a security best-practice violation as it creates a surface for argument or command injection if an attacker-controlled branch name or commit message is processed.
[INDIRECT_PROMPT_INJECTION]: The skill has a potential surface for indirect prompt injection as it processes data from the repository environment.
Ingestion points: Git branch names, file names, and commit history are read and processed in references/workflow-commit.md and references/workflow-pr.md.
Boundary markers: None; there are no delimiters or instructions used to prevent the agent from being influenced by maliciously crafted Git metadata.
Capability inventory: The skill has permission to modify the repository through git commit, git push, and gh pr create operations.
Sanitization: None; Git metadata is interpolated directly into shell templates and pull request descriptions without validation or escaping.

git