Workflow S2: Autonomous multi-round security review loop — a fresh-context Big-Four reviewer critiques experiments/results, the executor fixes, re-runs, and re-reviews until positive assessment or max rounds. Use when user says "auto review loop", "security…
Independent integrity audit of security experiment EVIDENCE (not code) before results become paper claims. A fresh-context reviewer checks oracle firing, ground-truth independence, dedup honesty, phantom results, authorization compliance, and scope…
Generate, stress-test, and pilot-validate cybersecurity, Web application security, and systems security research ideas for top venues (IEEE S&P, USENIX Security, ACM CCS, NDSS). Use when user says "网安找idea", "Web安全选题", "sec-idea-creator", "security research…
Workflow S1: Full security idea discovery pipeline — from a broad Web/systems-security direction to validated, pilot-tested, refined ideas with an experiment roadmap. Chains sec-research-lit → sec-idea-creator → sec-novelty-check → sec-research-review →…
Draft and audit cybersecurity, Web application security, and systems security papers for IEEE S&P, USENIX Security, ACM CCS, NDSS, PETS, and ACSAC. Enforces threat-model tables, adaptive-attacker evaluation, single fact-pack provenance, CVE/commit pinning,…
Workflow S0: Full end-to-end security research pipeline — broad direction → idea discovery → experiment bridge → auto review loop → narrative summary → optional paper writing. Use when user says "网安全流程", "从找idea到投稿", "security end-to-end research", or wants…
Turn a validated security idea (attack, defense, or measurement) into a threat-model-anchored, minimal-mechanism, implementation-ready proposal via iterative fresh reviewer rounds. Use after sec-idea-creator/sec-idea-discovery, or when the user says "细化安全方案",…
Judge what security claims the experiment evidence actually supports, what it doesn't, and what's missing — then route to confirm / supplement / pivot / negative-finding. Standalone version of the routing embedded in sec-experiment-bridge; rerun after every…
Design the ablation matrix for a security paper after main results pass sec-result-to-claim (supported or partial). Covers the four security ablation axes — component deletability, attacker tier (A1→A3), oracle coverage, and target-family diversity — with one…
Workflow S1.5: Bridge between security idea discovery and auto review. Reads EXPERIMENT_PLAN.md, implements the security harness (targets, oracles, A3 probes, baselines), deploys experiments on containerized/local infrastructure, collects oracle-verified…
Turn a refined security proposal into a claim-driven, oracle-first experiment roadmap: A1/A2/A3 attacker tiers, matched-budget baselines, ablation matrix, statistical protocol, ethics checkpoints, and artifact-ready run order for Big-Four submission. Use…
Verify root-cause, mechanism, and trust-boundary novelty for cybersecurity, Web application security, and systems security ideas against Big-Four papers, DBLP, RFCs, and public advisories/CVEs. Compatible with ARIS idea_discovery_gate.py.
Post-draft autonomous improvement loop for security manuscripts: fresh-context Big-Four review → targeted LaTeX fixes → recompile → re-review, for up to N rounds. Applies the same hard gates as sec-auto-review-loop to the *text layer* (threat-model…
One-shot pipeline chaining sec-research-refine (threat-model-anchored method refinement) and sec-experiment-plan (oracle-first, A3-aware experiment roadmap). Use when the user wants the full method+experiment package in one run, or asks to "串起来" the security…
Deep adversarial review of cybersecurity, Web application security, and systems security ideas, pilots, or manuscripts from a Big-Four PC perspective (IEEE S&P, USENIX Security, ACM CCS, NDSS). Audits threat-model sanity, TCB realism, ground-truth oracles, A3…