The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from an external source (Linear) and correctly identifies the potential for indirect prompt injection. It includes explicit defensive instructions to mitigate this risk.
Ingestion points: Data enters the agent context through Linear ticket lookups and searches (e.g., orca linear issue --current --full).
Boundary markers: The skill includes a dedicated 'Read First' section that mandates treating all returned Linear fields as untrusted source data and explicitly forbids the agent from following instructions found in ticket text, comments, or attachments.
Capability inventory: The skill can execute subprocesses via the orca CLI and read file content via the --body-file argument.
Sanitization: Uses instruction-based boundaries to ensure the agent uses ticket content only as reference material.
[COMMAND_EXECUTION]: The skill uses the orca and orca-ide CLI tools. These commands are scoped to project management tasks such as searching issues, updating status, and creating follow-up tickets. There is no evidence of arbitrary command execution.

linear-tickets