AI agent interoperability protocols enable orchestrators to discover and invoke external agents using machine-readable metadata. In the Agent2Agent (A2A) protocol, Agent Cards contain attacker-controllable natural-language fields that may influence agent selection before a provider is invoked, creating a pre-invocation trust problem. We characterize this discovery-time attack surface through a taxonomy of ten Agent Card manipulation attacks and present SECCARD, a layered validation framework combining protocol checks, deterministic analysis, and LLM-based semantic reasoning. We measure the emerging public A2A ecosystem by probing 36.1 million Internet domains and collecting 310 deployed Agent Cards. Two annotators independently label the corpus with 96.8% agreement. We then construct a 660-card real-card perturbation benchmark with matched benign controls and a 260-card synthetic adversarial benchmark spanning all ten attack categories. On controlled attacks injected into real Agent Cards, the strongest semantic judge achieves 97.0% security recall at a 1.7% false-positive rate. Across nine LLM judges on the synthetic benchmark, security recall ranges from 25.8% to 92.3%, demonstrating substantial model-dependent variation. These results show that SECCARD effectively identifies Agent Card manipulations spanning all ten discovery-time attack classes while preserving benign Agent Cards, but also reveal that its effectiveness depends critically on the choice of the underlying LLM judge.
Figures & tables
Work
A2A
Agent Card Security
Agent Card Attack Classes
Wild Agent Cards
Human Ground Truth
Pre-Invocation Defense
Habler et al. [ 12 ]
✓
✓
–
–
–
–
Louck et al. [ 13 ]
✓
–
–
–
–
–
Li et al. [ 16 ]
✓
✓
2
–
–
–
This work
✓
✓
10
✓
✓
✓
TABLE I: Comparison of related work on A2A and agent security.
A2A Card Fields
Type
Free-text?
Required fields
name
string
✓
description
string
✓
supportedInterfaces
object
×
version
string
×
capabilities
object
×
TABLE II: A2A agent card schema fields according to the v1.0 specification. The Free-text? column marks fields containing natural language parsed by an LLM consumer.
Fig. 1: A2A Discovery-to-Invocation Process. The Agent Card trust boundary is indicated by the red arrow.
Fig. 2: Taxonomy of Agent Card Manipulation attacks resulting from agent card poisoning. Attacks are grouped according to their primary objective.
ID
Attack
Analysis Type
Primary Detector
II
Indirect injection
Semantic
Layer 2B
RM
Ranking manipulation
Semantic
Layer 2B
SD
Skill deception
Semantic
Layer 2B
SC
Skill chaining
Semantic
Layer 2B
EM
Evaluator manipulation
Semantic
Layer 2B
OC
Out-of-context skills
Mixed
Layer 2A + Layer 2B
TABLE III: Relationship between discovery-time attacks and their primary detection mechanisms in SecCARD.
Fig. 3: SecCARD architecture for pre-invocation Agent Card validation. Layer 1 verifies structural and protocol conformance, while Layer 2 applies complementary deterministic (Layer 2A) and LLM-based semantic (Layer 2B) analyses. The resulting security verdict can inform a deployment-specific trust and admission policy.
Annotator 2
Annotator 1
Benign
Suspicious
Malicious
Benign
277
6
0
Suspicious
4
22
0
Malicious
0
0
1
TABLE IV: Inter-annotator agreement on the 310 in-the-wild Agent Cards.
Fig. 4: Excerpt from the description field of the emc2ai.io agent card. The highlighted sentence is a Ranking manipulation directive intended to influence agent selection.
Property
In-the-Wild
Real-Card Perturbation
Synthetic Adversarial
Total cards
310
660
260
Benign
282
60
0
Suspicious
27
300
85
Malicious
1
300
175
Source
Internet crawl
Real-card seeds
Fully generated
Human ground truth
✓
–
–
TABLE V: Overview of the three datasets used in our evaluation.
Model
Acc.
Recall
FPR
Prec.
κ3
GPT-5.5
94.5%
89.3%
5.0%
64.1%
0.718
Claude Opus 4.8
76.1%
96.4%
25.9%
27.0%
0.329
GPT-5.4-mini
82.9%
85.7%
17.0%
33.3%
0.394
Gemini 3.1 Flash Lite
73.2%
100.0%
28.4%
25.9%
0.292
Claude Haiku 4.5
57.4%
100.0%
45.7%
17.8%
0.162
Gemini 2.5 Flash Lite
91.3%
82.1%
0.7%
92.0%
0.463
TABLE VI: Judge performance on the 310 in-the-wild Agent Cards. Recall treats sus ∪ mal as positive ( n=28 ); FPR is computed over the 282 benign cards.
Closed-Source
Open-Source
Category
Group
N
GPT-5.5
Claude Opus 4.8
GPT-5.4 mini
Gemini 3.1 Flash Lite
Claude Haiku 4.5
Gemini 2.5 Flash Lite
Llama 3.1
Phi-4
Mistral Nemo
Params
—
—
—
—
—
—
8B
14B
12B
indirect_injection
Semantic
60
100%
100%
98%
98%
97%
97%
82%
67%
52%
ranking_manipulation
Semantic
60
100%
75%
52%
82%
65%
60%
62%
17%
23%
skill_deception
Semantic
60
98%
85%
95%
88%
75%
83%
62%
35%
13%
skill_chaining
Semantic
60
98%
92%
98%
95%
80%
78%
60%
25%
28%
TABLE VII: Cross-model security recall on the 660-card real-card perturbation dataset. The Controls block reports behaviour on the 60 untouched seed cards; seed accuracy and seed false-positive rate are complements, shown separately for readability.
Closed-Source
Open-Source
Category
Group
N
GPT-5.5
Claude Opus 4.8
GPT-5.4 mini
Gemini 3.1 Flash Lite
Claude Haiku 4.5
Gemini 2.5 Flash Lite
Llama 3.1
Phi-4
Mistral Nemo
Params
—
—
—
—
—
—
8B
14B
12B
indirect_injection
Semantic
35
100%
97%
97%
100%
91%
89%
80%
54%
43%
ranking_manipulation
Semantic
35
86%
57%
31%
69%
60%
57%
66%
17%
17%
skill_deception
Semantic
35
89%
69%
89%
86%
77%
71%
57%
40%
23%
skill_chaining
Semantic
35
100%
91%
91%
91%
89%
86%
66%
17%
20%
TABLE VIII: Cross-model security recall on the 260-card synthetic adversarial dataset. Security recall: MALICIOUS → MALICIOUS; SUSPICIOUS → {SUSPICIOUS,MALICIOUS}. There are no BENIGN cards, so this table admits no false-positive rate. “—” in the Params row denotes undisclosed model size.
Artificial intelligence agent security studies attacks and safeguards for tool-using, persistent, and collaborative systems across digital and physical settings.