The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill documentation in reference/pipeline.md provides shell commands using sudo apt-get install to set up the build environment on Linux, which requires elevated administrative privileges.\n- [EXTERNAL_DOWNLOADS]: The skill instructions involve cloning the PS2Recomp repository from GitHub and downloading related tools. These are standard development operations for the skill's domain.\n- [INDIRECT_PROMPT_INJECTION]: The skill parses untrusted binary data from local ISO and ELF files, representing a theoretical surface for indirect prompt injection.\n
Ingestion points: Binary data is read from user-provided local files via iso.py and elf.py.\n
Boundary markers: The instructions define strict behavioral constraints and a 'Circuit Breaker' mechanism to prevent unsafe automated behavior.\n
Capability inventory: The agent is authorized to write to the local filesystem and execute shell commands for building software.\n
Sanitization: The iso.py script includes a safe_component function to validate filenames and prevent path traversal during file extraction.\n- [COMMAND_EXECUTION]: The skill requires the agent to execute various command-line tools for project management, file processing, and software compilation.

console-reverse