The Agent Skills Directory

[PROMPT_INJECTION]: The instructions use extreme imperative language such as "EXTREMELY-IMPORTANT", "ABSOLUTELY MUST", and "NOT NEGOTIABLE" to enforce a mandatory skill-checking workflow. It explicitly commands the agent to suppress internal reasoning that might lead to skipping the check ("You cannot rationalize your way out of this", "These thoughts mean STOP"), which is a pattern characteristic of instruction overriding.
[PROMPT_INJECTION]: The skill establishes an indirect prompt injection surface by routing agent actions based on user-provided strings. This allows untrusted input to trigger the loading of specialist modules. 1. Ingestion points: User task descriptions provided at the start of the conversation (e.g., "Implement the auth endpoint"). 2. Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands within the user input used for routing. 3. Capability inventory: The skill orchestrates the loading of modules with significant capabilities, including browser interaction (browser-agent) and file/dependency modification (dependency-management). 4. Sanitization: No sanitization or validation logic is defined to prevent adversarial input from manipulating the routing process.

using-superantigravity