The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various subcommands of the devspace CLI, such as run, wait, show, and continue, to manage delegated tasks.
[INDIRECT_PROMPT_INJECTION]: The skill operates as a bridge for passing external data and user instructions to subagents, creating a surface for indirect prompt injection attacks.
Ingestion points: Untrusted data enters the subagent context via the <brief> argument in the devspace agents run and devspace agents continue commands defined in SKILL.md.
Boundary markers: The instructions do not specify any delimiters or safety headers to prevent the subagent from interpreting content within the brief as high-priority system instructions.
Capability inventory: The subagents have the ability to write to the local filesystem (enabled by default in allowed mode or explicitly via full_access) and perform git operations as described in SKILL.md.
Sanitization: There are no instructions for sanitizing or escaping the content of the briefs or the project data before they are processed by the subagents.

subagents