The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local git and gh CLI commands to manage repository branches, commits, and pull requests. It utilizes shell heredocs (cat <<'EOF') for commit messages and PR bodies, which is a security best practice that prevents shell expansion and mitigates command injection risks from variable content.
[DATA_EXFILTRATION]: Includes explicit safety guidelines for the git commit process, instructing the agent to monitor staged files and specifically avoid including sensitive files such as .env or credential files.
[PROMPT_INJECTION]: The skill processes untrusted data from Jira ticket summaries (via getJiraIssue) and user input. It identifies this as an indirect prompt injection surface and implements a cleaning phase to strip punctuation and prefixes. Evidence: (1) Ingestion points: Jira issue summaries and user-provided ticket descriptions. (2) Boundary markers: Absent. (3) Capability inventory: Subprocess execution of git and gh in SKILL.md. (4) Sanitization: A 'Cleaning the story name' logic is provided to normalize the string before interpolation.

gett-push