The Agent Skills Directory

[COMMAND_EXECUTION]: The skill repeatedly executes shell scripts located within the project's .guardrails/scripts/ directory.
Specifically, it calls task-worktree.sh for starting, merging, and discarding worktrees.
It also executes guidelines-file.sh to determine which guidelines apply to specific files.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a dispatcher that ingests external data to generate prompts for other agents, creating an attack surface for indirect prompt injection.
Ingestion points: Plan files (e.g., docs/plans/2026-08-30-<topic>.md), git diff output, and project-specific guidelines files.
Boundary markers: The instructions do not define delimiters or "ignore embedded instructions" warnings when passing data from the plan or diffs into the subagent's prompt.
Capability inventory: The skill has the ability to execute shell commands, perform git merge operations, and dispatch new subagents with custom instructions.
Sanitization: There is no mention of escaping, filtering, or validating the content extracted from plan files or code diffs before it is interpolated into subagent prompts.

develop-change