The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from external issue trackers which could influence agent behavior.
Ingestion points: Issue bodies and comments are retrieved via scripts/tickets.sh (subcommands view and next) and utilized throughout the execution spine in SKILL.md.
Boundary markers: Partial boundary instructions exist (e.g., prioritizing ## Agent Brief comments), but the skill lacks strict isolation or delimiters for external data.
Capability inventory: The skill possesses extensive capabilities, including file system modifications (git add, git commit), network operations (git push, GitHub/Linear/Jira API interactions), and execution of validation scripts.
Sanitization: No explicit sanitization or filtering of issue content is performed before the agent processes it to generate code.
[COMMAND_EXECUTION]: Core functionality relies on executing various shell commands and utility tools.
Invokes project-specific tools like git, gh, and orca as part of the implementation workflow.
Executes locally provided scripts (tickets.sh, open-pr.sh, text-frame.sh) to interface with trackers and manage pull requests.
[DYNAMIC_EXECUTION]: The skill uses dynamic script execution for data processing.
scripts/tickets.sh embeds and executes Python code blocks using python3 - to handle complex API interactions with Linear and Jira.

cast