- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
This paper introduces “AgentPhishing”, which is a benchmark for evaluating phishing detection effectiveness and how often agents comply with malicious requests under social-engineering attacks in email tasks. The authors also present “PhishOpt”, an attack method that adapts existing work “SkillOpt” to optimize reusable prompts for generating emails in a pipeline. The pipeline also uses Cialdini’s persuasion principles within the pipeline to label generated emails and use those to write a universal generation skill/prompt. The authors present experiments across six models used to develop the attacks and nine additional models used for evaluation. The work shows that some models handle more basic phishing attacks safely but fail on more tailored attacks.
Critical strengths
The paper addresses an important practical problem and usefully separates phishing detection from compliance with malicious requests. The results show why conventional phishing evaluations may give an incomplete picture of agent safety.
Critical weaknesses
The contribution over existing work seems narrow. PhishOpt adapts SkillOpt, while AgentBait already studies social engineering against agents and PEEK uses persuasion analysis for phishing generation. I’m not convinced that this work makes a substantial contribution. The paper focuses on phishing generation by combining existing methods and adapting the pipeline narrowly.
The evaluation lacks validation. Email realism and persuasion labels are assessed by LLMs without reported human validation. This leaves uncertainty about both the plausibility of the attacks and the reliability of the persuasion analysis because evaluating LLMs with imperfect LLMs leaves room for unreliable evaluation.
Existing adaptive methods are not used for comparison. The main baselines are static emails and an unoptimized prompt. These comparisons do not show that PhishOpt offers a meaningful improvement over other optimization approaches such as TextGrad.
Questions for authors
- What is novel in this work beyond combining AgentBait, SkillOpt, and PEEK and deriving a prompt based on including Cialdini’s persuasion principles?
- How does PhishOpt compare with an existing adaptive attack method or a generic prompt optimizer under the same generation and target-query budgets?
- What evidence validates the LLM realism judgments and Cialdini labels? I think the evaluation should be stronger than relying on LLMs that might themselves be imperfect.
3 comments
Thank you for your response and the extra experiments.
W1 (novelty): This remains my biggest concern. I still feel that the contribution is very narrow. PhishOpt is built on SkillOpt. Using a target’s feedback to improve an attack is also not new, as your own related-work section shows (AutoInject, MUZZLE, SpearBot). So the contribution still looks like existing techniques applied to a new setting, where considerable work already exists.
W2: I appreciate the added human validation. However, the human check relies on a single annotator, so there is nobody to compare their labels against. I think two or more independent annotators are needed for the check to be convincing. I also noticed that the annotator rated 5 of the 28 emails below your realism threshold, compared with 1 for the LLM judge, so the judge may be more lenient than a human.
W3: Thank you for adding this baseline. As you note, the TextGrad comparison covers one agent and one run per method. The gap between PhishOpt and TextGrad inside your pipeline seems small compared with the variation between batches. That is too little to conclude that PhishOpt is better. Results on more agents, with several runs per method, would make the comparison convincing. It may also be helpful to report results with unmodified SkillOpt.
I am keeping my rating.
W1(Novelty): Our contribution is not the optimization mechanism but the threat model it measures and the findings it produces: what a tool-using email agent does with a persuasive but non-injecting message inside a task loop. To our knowledge, no prior work measures this quantity. PhishOpt introduces a phishing-specific objective (detection and compliance), a realism gate that enforces persuasion-only generation, locked synthetic-only generation constraints, and a universal skill built from persuasion analysis.
| Method | What it optimizes | Feedback signal | Target | Channel |
|---|---|---|---|---|
| PEEK (Chen et al., 2024) | phishing generator | real-vs-synthetic discriminator | phishing classifier | persuasive email |
| SpearBot (Qi et al., 2025) | individual emails | detector critiques | phishing detector | persuasive email |
| AutoDAN-Turbo (Liu et al., 2025) | jailbreak strategy library | target responses | LLM (jailbreak) | jailbreak prompt |
| RL-Hammer (Wen et al., 2025) | attacker-model weights (RL) | black-box target reward | LLM agent | prompt injection |
| AutoInject (Chen et al., 2026) | injection suffixes (RL) | target reward | LLM agent | prompt injection |
| MUZZLE (Syros et al., 2026) | injection location + payload | execution trajectory | web agent | prompt injection |
| PhishOpt (ours) | natural-language skill | agent detection + compliance | email agent in a task loop | persuasion only (realism-gated) |
W2: We added 2 more annotators scored against the LLM judges. With realism, all three rated the generated emails as realistic with means around 0.81-0.91 and the judge placed both degraded controls below 0.7. With persuasion, one annotator agreed strongly with the key (mean κ = 0.62), a second moderately (κ = 0.30), and a third weakly (κ = 0.11). Agreement was consistent on scarcity/urgency (κ = 0.48–0.62 across all three) and dropped on the remaining principles for the two weaker annotators. This bounds the precision of the fine-grained persuasion coding but not the behavioral findings, which are computed from agent actions rather than these labels.
| Realism | Answer key (LLM) | Annotator A | Annotator B | Annotator C |
|---|---|---|---|---|
| Generated emails, mean | 0.90 | 0.91 | 0.81 | 0.82 |
| Degraded controls, mean | 0.33 | 0.81 | 0.28 | 0.84 |
| Controls below the 0.7 gate | 2/2 | 1/2 | 2/2 | 0/2 |
| Generated emails below the gate | 1/28 | 2/28 | 5/28 | 1/28 |
| Principle | Key present | Annotator A | Annotator B | Annotator C |
|---|---|---|---|---|
| Authority | 1.00 | —* | —* | —* |
| Scarcity / Urgency | 0.75 | 0.48 | 0.50 | 0.62 |
| Commitment & Consistency | 0.79 | 0.48 | 0.73 | 0.00 |
| Social proof | 0.14 | 0.18 | 0.51 | −0.24 |
| Liking | 0.54 | 0.11 | 0.50 | 0.14 |
| Reciprocity | 0.14 | 0.28 | 0.84 | 0.13 |
| Mean κ | 0.30 | 0.62 | 0.11 |
* Authority: the key marks it present on all 28 emails, so κ is undefined (no variance); raw agreement is 82–100%.
W3: We have extended the experiment to five agents and split it into two. Experiment 1: is PhishOpt better than a generic adaptive optimizer? Experiment 2: does the advantage come from the editor or the loop?
Experiment 1: We hold the base skill, generator, engine model, feedback, and budge fixed. We ran both methods against five target agents. From the table below, we can see that PhishOpt exceeds standard TextGrad on every agent suceptile attacks on Qwen, GPT, Grok. There are two exceptions with Gemini and Sonnet, where they are robust to all methods.
| Target agent | Base | TextGrad | PhishOpt |
|---|---|---|---|
| qwen3.7-plus | 0.090 | 0.148 | 0.519 |
| gpt-5.4 | 0.062 | 0.112 | 0.243 |
| grok-4.20 | 0.090 | 0.139 | 0.252 |
| gemini-3.5-flash | 0.069 | 0.074 | 0.024 |
| claude-sonnet-5 | 0.035 | 0.043 | 0.037 |
Experiment 2: To determine whether PhishOpt's advantage comes from its editor or from the surrounding loop (the realism gate and best-so-far selection), we held the loop fixed and replaced only the editor with TextGrad's. When TextGrad is given PhishOpt's loop, its performance rises sharply, showing that the loop is a core strength of our approach. PhishOpt's editor performed well on Qwen and GPT, while on Grok, TextGrad-in-loop scored higher. This does not reflect a more effective attack: the gain came from the agent incorrectly flagging legitimate emails as phishing without causing it to act on a genuine phishing request. An independent judge also rated the realism of every email, confirming that no method improved its score by reducing realism.
| Target agent | TextGrad-in-loop | PhishOpt |
|---|---|---|
| qwen3.7-plus | 0.150 | 0.519 |
| gpt-5.4 | 0.170 | 0.243 |
| grok-4.20 | 0.313 | 0.252 |
| gemini-3.5-flash | 0.108 | 0.024 |
| claude-sonnet-5 | 0.044 | 0.037 |
W1: We thank the reviewer for asking us to clarify our contribution. Prior work like PEEK evolves phishing emails against a classifier to train phishing detectors, and its code is not publicly available. AgentBait studies social engineering against agents, but its attacks are predefined and fixed before the agents are observed. PhishOpt is the first method that adapts attack generation to the evaluated agent, using an objective score combining detection, compliance, and F1 as feedback.
To address concerns between PhishOpt and similar Phishing detection methods, we compared directly with AgentBait predefined attacks. We converted 73 released email-scenario attacks to text and evaluated them against seven agents. From the table, AgentBait's attacks are about as detectable as our base skill, while PhishOpt evades detection on six of the seven agents.
PhishOpt = each agent's optimized skill (original 5), universal skill (held-out 2).
W2(Regarding human validation on realism and persuasion labels): We agree that LLM judgement without human validation leaves uncertaintity. To validate both judges, one human annotator hand labbeled a sample of 30 emails for realism and persuaion principles, using the same criteria used by the LLMS (table below). For persuasion principles, LLM and human agreed on 89% of the time with the strongest agreement on commitment and consistency. For realism, both rate generated emails as realistic and both degraded on emails below the 0.7 gate.
Realism, LLM judge vs human (30 items; human 1–5 mapped to 0–1):
W3: We thank the reviewer for this suggestion. To address this concern, we ran TextGrad on GPT-5.4 with the same base skill, generator, engine, model, feedback, budget in two configurations. One where we use standard textGrad and the other where we use TextGrad inside our pipeline where we replace phishOpt editor with TextGrad Update step and keep everything else the same. We evaluated each final skill on three fresh batches (table below).
From the table, PhishOpt scored highest on every test batch and substantially reduced detection rate without raising false positives. TextGrad in the PhishOpt pipeline increased in score but it mainly came from GPT-5.4 flagging more benign emails which an increase in false positive. This comparison covers one target and one run per method, and TextGrad used more engine calls than PhishOpt (three per update vs. one); we will extend it to additional targets in the revision.