The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by ingesting untrusted data from external GitHub issues which may contain malicious instructions designed to alter the agent's behavior during the planning phase.
Ingestion points: The workflow in SKILL.md fetches content (title, body, and comments) from GitHub issues using the gh issue view command and identifies this as the "core problem statement" to guide investigation.
Boundary markers: There are no delimiters or boundary markers used to insulate the agent's logic from potentially adversarial content in the issue body.
Capability inventory: The skill possesses significant capabilities including the ability to search and read the entire local codebase (code_search, read_file) and write content back to the remote repository (gh issue comment).
Sanitization: There is no evidence of sanitization, filtering, or validation of the external issue data before it is used to determine the investigation focus and implementation steps.
[COMMAND_EXECUTION]: The skill executes shell commands using the GitHub CLI to interact with external repositories.
Evidence: SKILL.md contains executable blocks for gh issue view, gh repo view, and gh issue comment. These commands are integral to the skill's stated purpose of issue management.

yds-gh-issue-planner