The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands including ls -La and git rev-parse to navigate task-specific directories and determine the current repository state (e.g., identifying if the agent is operating within a git worktree). These commands are used for legitimate project management and state verification.
[INDIRECT_PROMPT_INJECTION]: The skill ingests research documents and task data from the .humanlayer/tasks/ directory to generate its output. While this input is technically untrusted, the skill's primary purpose is to synthesize this information into a markdown document for human review. The risk of unintended instruction execution is mitigated by the review-centric workflow and the standard nature of the data being processed.

create-structure-outline