The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection via GitHub Issues.
Ingestion points: Untrusted data enters the agent context via the gh issue view command in SKILL.md.
Boundary markers: Absent. The skill does not use delimiters or warnings to separate external issue content from its instructions.
Capability inventory: The skill employs Bash, Edit, Write, and Agent tools (SKILL.md) to modify the codebase and execute commands like go test.
Sanitization: Absent. No filtering or escaping is applied to the fetched issue content before it is processed.
[COMMAND_EXECUTION]: Automated execution of logic derived from external content. The skill executes go test and go build on code that is implemented based on requirements found in GitHub issues. While this is the intended behavior, it constitutes a surface for executing logic influenced by untrusted external actors.

start-issue