The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes manifest files which are external inputs that could contain malicious instructions aimed at influencing the ticket generation process.
Ingestion points: Reads manifest files from ~/.manifest-dev/manifests/ or other user-provided paths.
Boundary markers: The instructions provide boundary markers by directing the agent to "translate, don't excerpt" and specifically to rewrite content in the ticket's own words rather than copying it directly.
Capability inventory: The skill possesses the capability to write files to the project root (within the tickets/ directory) and interact with GitHub via the gh CLI, API, or MCP tools to create issues and labels.
Sanitization: The transformation logic serves as a sanitization step, requiring the agent to strip manifest-dev vocabulary, verification codes, and ASM codes, converting them into plain-prose rules.
[COMMAND_EXECUTION]: The skill instructions include the use of external tools such as the gh CLI to interact with GitHub repositories for ticket management.
The interaction is scoped to managing issues, labels, and tracking relations based on the user's repository access.

ticket-up