The Agent Skills Directory

PROMPT_INJECTION (HIGH): The 'BANNED BULLSHIT Protocol' implements a behavior override that aggressively prohibits standard AI response patterns (e.g., 'AI Preamble Slop') and safety formatting. The use of hyperbolic language like 'capital offense' and 'kill it' is typical of jailbreak-style prompt injection intended to bypass model constraints.
COMMAND_EXECUTION (HIGH): The skill is designed to run multiple unverified Bash scripts (detect-stack.sh, scan-structure.sh, etc.) that perform deep scans of the local filesystem. This creates a high-risk surface for Indirect Prompt Injection (Category 8) where malicious content within a codebase could trigger unauthorized agent actions. Evidence Chain: 1. Ingestion points: Files are discovered and read via scripts using find, grep, and wc across the entire project directory. 2. Boundary markers: Absent; there are no instructions or delimiters to prevent the agent from obeying instructions embedded in the analyzed code. 3. Capability inventory: The skill has full access to the Bash tool, allowing for arbitrary command execution. 4. Sanitization: No validation or escaping of the codebase content is performed before processing.

codebase-context