The Agent Skills Directory

[PROMPT_INJECTION]: The skill has an indirect prompt injection surface due to its core functionality of processing external, untrusted data.
Ingestion points: The skill fetches issue and pull request titles, bodies, and comments from GitHub using gh issue list and gh pr list as shown in scripts/git-triage.sh.
Boundary markers: There are no instructions in the skill's execution flow to wrap these external strings in delimiters or to ignore any instructions embedded within them.
Capability inventory: The skill enables the agent to perform administrative actions including closing issues (gh issue close) and merging pull requests (gh pr merge).
Sanitization: The fetched text is provided to the agent without specific sanitization or filtering for adversarial prompt content.
[COMMAND_EXECUTION]: The skill executes shell commands to perform repository management tasks.
It utilizes the GitHub CLI for operations such as gh issue close and gh pr merge to update the repository state.
Risk Mitigation: These write operations are explicitly guarded by an AskUserQuestion step, ensuring a human-in-the-loop confirmation before any changes are committed.
[COMMAND_EXECUTION]: The skill implements dynamic context injection to enhance the agent's situational awareness.
It uses the ! syntax in SKILL.md to execute git remote -v, git rev-parse, and git branch commands at the time the skill is loaded.
These commands are benign and are standard practices for development tools to identify the current project context and branch status.

git-triage