The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to facilitate the git development lifecycle and GitHub integration.
Evidence: Commands such as git rev-parse to capture SHAs, git commit for checkpointing state, and git push -u origin HEAD to transmit work are core to the lifecycle.
Safety Context: The skill's doctrine explicitly prohibits force-pushing and modifications to the default branch, and it requires authentication preflights before pushing.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Linear tickets, creating a vector for potentially malicious input to influence sub-agents.
Ingestion points: Ticket details are resolved using the /trackers command as defined in doctrine.md and reference.md.
Boundary markers: The skill employs a visible 'Execution context' in the chat that strictly defines binary 'Done when' checks, 'non-goals', and specific 'Active Rules' to constrain behavior.
Capability inventory: The agent has filesystem write access via git, the ability to invoke sub-agents like /task for implementation, and network write capabilities for PR creation via gh or Cursor tools.
Sanitization: Although explicit string sanitization of ticket content is not detailed, the rigid transition between research, implementation, and two mandatory code review phases (CR1/CR2) ensures implementation is validated against ticket intent before shipping.

just-do-it