Decepticon

AATMF T5 — Model & API Exploitation. Rate-limit abuse, token-cost amplification, schema bypass, model-version manipulation.

AATMF T4 — Multi-Turn & Memory Manipulation. Persistent memory injection, conversation-state poisoning, cross-session contamination, ghost-context leak.

AATMF T3 — Reasoning & Constraint Exploitation. System prompt override, constraint negation, role-reversal, instruction conflict exploit.

AATMF T2 — Semantic & Linguistic Evasion. Foreign-language pivot, encoded payloads, esolang, fictional framing, jailbreak via translation.

AATMF T1 — Prompt & Context Subversion. Direct + indirect prompt injection, ASCII smuggling, payload-in-image, prompt-leaking via reflection.

LLM red team category — full AATMF v3 tactic coverage (T01–T15). Routing skill: read this first to identify which tactic applies, then load the matching sub-skill. Maps to MITRE ATLAS where overlap exists.

Stage 5 exploit construction playbook. Weaponizes validated primitives into multi-step chains that reach crown jewels. Load at exploiter-agent startup.

Stage 2 vulnerability detector playbook. Reads source around CANDIDATE nodes and promotes real bugs to VULNERABILITY + HYPOTHESIS. Read-only. Load at detector-agent startup.

Benchmark mode marker — engagement objective is flag capture. Generic engagement rules apply unchanged.

Drive Decepticon — an autonomous multi-agent red-team framework — over MCP to run authorized penetration tests and bug-bounty engagements end to end, then watch and steer them live from chat. Launch an engagement against a target, poll its transcript to narrate progress, send messages to refocus it, and pull findings as SARIF. Use when the user asks to run a pentest/red-team engagement, hunt a bug bounty, do recon, exploit/scan a host, web app, API, network, cloud, Active Directory, mobile app, or smart contract WITH Decepticon — or to check/resume a running engagement or report what Decepticon found. Triggers: run a decepticon engagement, pentest this with decepticon, bug bounty, recon this target, red team this, scan this host, resume the engagement, what did decepticon find, decepticon status. Do NOT use for ad-hoc local tool runs (running nmap/sqlmap/ffuf directly) when no Decepticon server is involved — this drives the Decepticon orchestrator, not raw tools.