The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external issue trackers (GitHub issues, Linear/Jira tickets). It explicitly mitigates the risk of indirect prompt injection by instructing the agent to treat all content from untrusted documents and tool outputs as data only, explicitly forbidding such content from authorizing actions.
[COMMAND_EXECUTION]: The skill utilizes a local shell script (scripts/tickets.sh) to interact with the gh and git CLIs. These are standard operations required for the skill's core functionality of managing tickets and repository state.
[DYNAMIC_EXECUTION]: The scripts/tickets.sh file contains an embedded Python adapter that is executed dynamically via a heredoc. This is an internal implementation detail for cross-platform compatibility with Linear and Jira APIs and does not involve executing externally fetched code.
[DATA_EXFILTRATION]: The skill communicates with well-known services (GitHub, Linear, Jira) using user-provided credentials via environment variables. The network operations are limited to the intended project management APIs, and the code includes security checks to ensure HTTPS is used and redirects remain on the same origin.
[CREDENTIALS_UNSAFE]: No hardcoded credentials were found. The skill properly relies on environment variables (e.g., LINEAR_API_KEY, JIRA_API_TOKEN) for authentication, following standard security practices.

conjure