The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from an issue tracking database, creating a vulnerability to indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through forge issue show, forge issue list, and forge issue search commands in SKILL.md.
Boundary markers: The skill does not provide clear delimiters or instructions for the agent to separate issue data from its operating instructions.
Capability inventory: The skill possesses write capabilities through forge issue update, forge issue create, and forge issue comment in SKILL.md, allowing injected instructions to potentially modify records.
Sanitization: No mechanism for sanitizing or validating retrieved text content from issue fields is described in the instructions.
[COMMAND_EXECUTION]: The skill relies on the execution of shell commands using the forge CLI for all issue management tasks, which is the primary mechanism of the skill.

issue-basics