The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the herdr command-line interface to manage the lifecycle of agents and workspace panes. Key operations include herdr pane split for UI organization, herdr agent start for initialization, and herdr agent prompt for inter-agent communication.
[EXTERNAL_DOWNLOADS]: Instructions within the playbooks (specifically for Architect and Lead roles) recommend the installation of additional tools using npx skills add. These references include resources from the vendor's own repository (adminoryslabs/Armory) and a third-party repository (tt-a1i/archify).
[INDIRECT_PROMPT_INJECTION]: The skill acts as a conduit for project documentation, such as PRDs and technical specifications, which are ingested from the workspace and passed to other agents. This creates a multi-step chain where malicious instructions embedded in a document could influence downstream agent behavior.
Ingestion points: Processes files like PRD.md, SPRINT0.md, and TECH-DESIGN.md found in the user's workspace.
Boundary markers: Relies on structured Markdown templates and role-specific checklists to separate instructions from data.
Capability inventory: Includes the ability to execute shell commands, manage agents via herdr, and interact with version control via git and gh (e.g., creating Pull Requests).
Sanitization: There is no evidence of automated sanitization or escaping of the content ingested from Markdown files before it is interpolated into prompts for other agents.

invoke-agent