The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from synced Hamster briefs, specifications, and tasks which are then interpreted by sub-agents with code-writing and shell execution capabilities.
Ingestion points: Markdown files within the .hamster/ directory (e.g., brief.md, spec.md, and task files).
Boundary markers: The instructions explicitly designate brief.md and spec.md as higher authority than individual tasks to mitigate conflicting instructions.
Capability inventory: The skill can execute shell commands (git, gh, hamster), write or modify codebase files, and run project-defined validation or test scripts.
Sanitization: There is no evidence of explicit content sanitization or escaping for the markdown data before it is provided to the task-executor sub-agent.
[COMMAND_EXECUTION]: The skill relies on extensive shell command execution to manage the development lifecycle.
Uses the hamster CLI for status updates, authentication, and synchronization.
Uses the GitHub CLI (gh) for repository information and pull request management.
Executes project-specific build and test tools such as npm, pnpm, yarn, cargo, go, and make based on detected configuration files.
[DYNAMIC_EXECUTION]: The skill uses Python and shell scripts to process logic and configuration at runtime.
Executes a Python script via a heredoc in references/brief-selection.md to parse and validate .hamster/.state.json.
Executes bundled shell (ensure-ready.sh) and PowerShell (ensure-ready.ps1) scripts to verify environment readiness.
[PRIVILEGE_ESCALATION]: Documentation in SKILL.md advises the user to run readiness commands with elevated permissions (sudo) if the platform's execution environment restricts access to system keychain credentials.

ship