| name | attack-selection |
| description | How to map an operator intent to one of the attacks discovered on the attacker host. Use this before calling execute_attack. |
Attack Selection
The attacker host exposes a catalog of attacks discovered at startup (each one is a directory under attacker_attacks_root with a README.md, Dockerfile, and entrypoint.sh). Your job is to pick the right one for the intent.
Tools
list_available_attacks() -> list[{attack_id, description, required_arguments}] — quick browse of every available attack.read_attack_definition(attack_id) -> {attack_id, description, docker_image, required_arguments, readme_excerpt} — full definition of a single attack including the README excerpt.
Process
- Always call
list_available_attacks first to confirm the catalog.
- Match the operator intent to the closest
attack_id based on description and required arguments. Mention IPs, ports, protocols, and attack class in your reasoning.
- Call
read_attack_definition(attack_id) for the candidate to inspect the README and required argument names.
- If no candidate is a defensible match, choose the closest one anyway and explain the mismatch in
evasion_rationale. The deterministic guardrail in the agent will raise if the attack_id is unknown — never invent an attack_id.
Argument Ordering
The required_arguments field returns a list of NAMES (e.g. ["target_ip", "target_port", "rate"]). You must produce a list of VALUES in the same order. Wrong order silently breaks the attack.
Multiple Reasonable Candidates
If two attacks could match the intent (e.g. both target the same protocol but differ in payload), choose the one whose required_arguments are best determined by the intent. Explain the choice in evasion_rationale.
