The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub issues via the gh issue view command in Step 1. This content is used to generate implementation plans and code, which could lead to the agent following malicious instructions embedded in the issue body.
Ingestion points: GitHub issue content fetched via gh issue view or provided directly by the user (SKILL.md, Step 1).
Boundary markers: No specific delimiters are used when the agent processes the issue content, though the output plan is isolated in the _/local-plans/ directory.
Capability inventory: The skill can execute shell commands (gh), run project tests (Step 6), and create commits via the create-commit skill (Step 7).
Sanitization: The instructions include defensive measures against shell injection by mandating the use of double quotes "${var}" and HEREDOC <<'EOF' when passing variables to the gh CLI.
[COMMAND_EXECUTION]: The skill performs several command-line operations, including fetching data from GitHub, searching for existing issues, and executing the project's test suite. While these are necessary for the development workflow, executing arbitrary project tests (Step 6) carries inherent risk if the local environment is untrusted.

implement-issue