The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes 'Ticket Envelope' artifacts and ticket bodies which are untrusted external inputs used to guide implementation and test creation.
Ingestion points: The SKILL.md file and references/skills-only.md describe the consumption and validation of ticket envelopes and markdown bodies.
Boundary markers: The instructions utilize 'normalized envelopes' but do not define specific text delimiters or 'ignore' instructions for the ticket body to prevent adversarial subversion.
Capability inventory: The skill has the authority to implement ticket scope (file editing), execute tests, and perform limited Git operations (merge/sync when authorized).
Sanitization: The skill relies on external validators like to-spec and ticket_contract.py for normalization, but does not specify internal sanitization of the input content.
[COMMAND_EXECUTION]: The skill instructions involve calling functions from external packages and executing utility tools to manage the ticket lifecycle.
Evidence: references/skills-only.md directs the agent to call functions from ticket-autopilot/scripts/autopilot/ticket_contract.py and utilize tools such as to-spec, to-tickets, explain-pr, and sync-local-pi for artifact processing and environment synchronization.

execute-ticket