The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes external content from user-provided references such as file paths, URLs, or issue numbers. This creates a surface for indirect prompt injection where malicious instructions embedded in project specifications or issue comments could influence the generated task descriptions or lead to unauthorized actions.
Ingestion points: External references, URLs, and file bodies fetched during the context gathering phase (SKILL.md).
Boundary markers: The skill uses structured markdown templates (<local-ticket-template>, <issue-template>) for its output, which provides some structure but does not include explicit instructions to ignore directives found within the ingested data.
Capability inventory: The skill can write files to the local .scratch directory and make API calls to external issue trackers (GitHub, Linear) to create or link issues.
Sanitization: There is no evidence of sanitization or filtering of the fetched external content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill instructs the agent to explore the codebase to understand the project domain and architecture, which typically involves the use of file system tools and directory traversal.
[EXTERNAL_DOWNLOADS]: The skill connects to configured issue trackers (such as GitHub or Linear) to publish and link tickets. These are well-known services used for their intended purpose in this context.

to-tickets