The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run local utility scripts and CLI tools to gather data and monitor status. Specifically, it mentions python TOOLS/usage_report.py for analyzing session history and TOOLS/Wait-External.ps1 (which uses the GitHub CLI gh) to monitor external CI runs.
[EXTERNAL_DOWNLOADS]: The documentation in references/shell-output-compression.md refers to an installer mechanism that fetches catalog tags and validates candidate versions of the RTK utility. This indicates the presence of a lifecycle management system for the skill's dependencies.
[DYNAMIC_EXECUTION]: The skill utilizes a local shell hook (TOOLS/hooks/rtk_safe_hook.py) to dynamically intercept and rewrite the output of specific shell commands like git status, pytest, and cargo test. This is intended to reduce token usage by compressing human-facing output while preserving raw output for machine-readable tasks.
[INDIRECT_PROMPT_INJECTION]: The process of rewriting shell output introduces a surface where external command results (like test failures or git logs) are interpreted and modified before reaching the agent.
Ingestion points: Shell command outputs from git, pytest, cargo, and go test processed by the RTK safe hook.
Boundary markers: The instructions explicitly define a 'safe set' of commands and warn the agent not to filter input intended for parsers or hash checks.
Capability inventory: The skill includes Python scripts, PowerShell scripts, and shell command execution via the gh CLI.
Sanitization: The skill advocates for a specific hook parser with regression tests to ensure the correctness of the shell output transformation.

token-efficiency