The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, creating a potential attack surface for indirect prompt injection.
Ingestion points: The skill retrieves external content from GitHub issues via the gh issue view command and reads findings from research.md, which is derived from external data.
Boundary markers: The instructions do not mandate the use of delimiters (e.g., XML tags or clear separators) when interpolating the content of issues or research results into the final plan.md output.
Capability inventory: The skill possesses Bash access for command execution and Write/Edit permissions for file system modification, which could be misused if malicious instructions in an issue are obeyed.
Sanitization: There are no explicit steps provided for sanitizing or escaping instructions that may be embedded within the issue description or comments.
[COMMAND_EXECUTION]: The skill uses the Bash tool to interact with the project environment and external services.
Evidence: The skill instructions explicitly call for the use of gh issue view to retrieve issue details. It also describes a workflow for executing lint and type check commands as part of the implementation plan's Definition of Done (DoD) verification process.

plan