The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from several local sources including docs/consistency.md, AGENTS.md, and the output of the wt config show command. This content is used to shape TaskDocuments and matrix profiles. There are no explicit boundary markers or sanitization steps defined to prevent malicious instructions within these files from influencing the agent's logic or subsequent command generation.
Ingestion points: docs/consistency.md, .wt/execution/tasks, and wt config show output.
Capability inventory: The agent has the ability to write files to .wt/config/profiles/, .wt/execution/tasks/, and execute shell commands via the wt binary.
Boundary markers: Absent.
Sanitization: Not specified.
[COMMAND_EXECUTION]: The skill relies on executing a local binary located at ./target/debug/wt as well as standard utilities like git, find, and sed. While these are characteristic of a development environment skill, it assumes the wt binary is trusted and requires the agent to have execution permissions in the repository's debug directory.

variants