OWASP Top 10 for Agentic Applications
The OWASP Top 10 for Agentic Applications (published December 2025, ids ASI01 to ASI10) covers software where a model plans and acts with tools. Each entry below lists the Keryn rules that carry it in their metadata (rules 0.8.0) and the edges they inspect.
This is an edge map, not a score. A count says how many rules look at a class of weakness and how sure they are, not how safe a system is; a class with few rules may simply need few, and no number here means a framework is satisfied. Confidence: P0 exploitable as written, P1 dangerous in context, P2 heuristic or posture. Edges are the relationships the rules inspect (reaches, calls, reads, writes, sends, grants, depends_on, trusts).
Coverage matrix
Section titled “Coverage matrix”| Entry | Rules | P0 | P1 | P2 | Edges |
|---|---|---|---|---|---|
| ASI01 Agent Goal Hijack | 216 | 6 | 137 | 73 | reaches 71 · calls 50 · reads 15 · writes 10 · sends 13 · grants 29 · depends_on 10 · trusts 17 · none 1 |
| ASI02 Tool Misuse and Exploitation | 497 | 33 | 290 | 174 | reaches 64 · calls 175 · reads 40 · writes 22 · sends 16 · grants 154 · depends_on 17 · trusts 7 · none 2 |
| ASI03 Identity and Privilege Abuse | 355 | 27 | 197 | 131 | reaches 38 · calls 44 · reads 25 · writes 16 · sends 22 · grants 188 · depends_on 5 · trusts 17 |
| ASI04 Agentic Supply Chain Vulnerabilities | 165 | 23 | 96 | 46 | reaches 12 · calls 28 · reads 3 · writes 1 · sends 16 · grants 19 · depends_on 84 · trusts 2 |
| ASI05 Unexpected Code Execution | 249 | 65 | 129 | 55 | reaches 8 · calls 157 · reads 1 · writes 9 · sends 16 · grants 37 · depends_on 21 |
| ASI06 Memory and Context Poisoning | 166 | 2 | 85 | 79 | reaches 13 · calls 26 · reads 4 · writes 33 · sends 12 · grants 65 · depends_on 9 · trusts 4 |
| ASI07 Insecure Inter-Agent Communication | 57 | 2 | 35 | 20 | reaches 14 · calls 9 · sends 6 · grants 17 · depends_on 1 · trusts 10 |
| ASI08 Cascading Failures | 84 | 0 | 39 | 45 | reaches 7 · calls 12 · reads 4 · writes 6 · sends 4 · grants 48 · depends_on 1 · none 2 |
| ASI09 Human-Agent Trust Exploitation | 59 | 3 | 36 | 20 | calls 15 · reads 4 · writes 1 · sends 1 · grants 26 · trusts 9 · none 3 |
| ASI10 Rogue Agents | 65 | 14 | 28 | 23 | reaches 6 · calls 5 · reads 4 · writes 11 · sends 8 · grants 27 · depends_on 4 |
ASI01 Agent Goal Hijack
Section titled “ASI01 Agent Goal Hijack”216 rules · confidence P0 6, P1 137, P2 73 · severity error 62, warning 125, info 29
Edges inspected: reaches 71 · calls 50 · reads 15 · writes 10 · sends 13 · grants 29 · depends_on 10 · trusts 17 · none 1
30 of 216 rules, surest first
- seceng.agentrules.secrets-to-url-command
- seceng.ciagent.agent-pr-head-checkout-on-pull-request-target
- seceng.mcp.py.tool-description-hidden-unicode
- seceng.mcp.ts.tool-description-hidden-unicode
- seceng.eval.public-dashboard-no-auth
- seceng.llm02-system-prompt-in-client
- keryn.browseragent.model-output-to-page-navigation-js
- keryn.browseragent.model-output-to-page-navigation-py
- keryn.outbound.model-output-as-message-recipient-js
- keryn.outbound.model-output-as-message-recipient-py
- seceng.agentrules.base64-decode-exec
- seceng.agentrules.bidi-override-character
- seceng.agentrules.invisible-unicode-character
- seceng.aiconfig.bidi-override-character
- seceng.aiconfig.escaped-invisible-unicode
- seceng.aiconfig.invisible-unicode-character
- seceng.asi01.agent-instructions-loaded-at-runtime-python
- seceng.asi01.agent-instructions-loaded-at-runtime-ts
- seceng.asi01.system-prompt-from-retrieved-or-tool-content-python
- seceng.asi01.system-prompt-from-retrieved-or-tool-content-ts
- seceng.asi06.memory-read-into-system-prompt-python
- seceng.asi06.memory-read-into-system-prompt-ts
- seceng.asi07.peer-agent-message-as-instructions-python
- seceng.asi07.peer-agent-message-as-instructions-ts
- seceng.ciagent.agent-prompt-interpolates-event-text
- seceng.ciagent.agent-step-env-secrets
- seceng.ciagent.agent-step-on-pull-request-target
- seceng.llm01.prompt-injection-taint
- seceng.llm01.taint.tool-result-to-tool-args-python
- seceng.llm04.js-ts-model-output-to-system-shell
ASI02 Tool Misuse and Exploitation
Section titled “ASI02 Tool Misuse and Exploitation”497 rules · confidence P0 33, P1 290, P2 174 · severity error 151, warning 291, info 55
Edges inspected: reaches 64 · calls 175 · reads 40 · writes 22 · sends 16 · grants 154 · depends_on 17 · trusts 7 · none 2
30 of 497 rules, surest first
- keryn.browseragent.model-output-to-page-script-js
- keryn.browseragent.model-output-to-page-script-py
- keryn.outbound.model-output-as-payment-amount-js
- keryn.outbound.model-output-as-payment-amount-py
- keryn.rustagent.model-output-to-process-command-rs
- keryn.skillcred.browser-credential-store-read
- keryn.skillcred.cloud-metadata-endpoint-access
- keryn.skillcred.macos-keychain-secret-read
- keryn.skillcred.os-credential-database-access
- keryn.skillcred.secret-manager-or-keyring-export
- keryn.skillcred.ssh-key-material-read
- keryn.skillpersist.destructive-filesystem-command
- seceng.llm05.taint.model-output-to-async-subprocess-python
- seceng.llm05.taint.model-output-to-browser-eval-js
- seceng.llm05.taint.model-output-to-child-process-js
- seceng.llm05.taint.model-output-to-code-interpreter-tool-python
- seceng.llm05.taint.model-output-to-os-command-python
- seceng.llm05.taint.model-output-to-remote-command-python
- seceng.llm05.taint.model-output-to-shell-quote-missing-js
- seceng.llm05.taint.model-output-to-sql-js
- seceng.llm05.taint.model-output-to-sql-python
- seceng.mcp.csharp.tool-process-start
- seceng.mcp.csharp.tool-sql-command-interpolated
- seceng.mcp.go.tool-exec-command
- seceng.mcp.go.tool-exec-shell-c
- seceng.mcp.go.tool-sql-string-built
- seceng.mcp.java.tool-jdbc-statement-concat
- seceng.mcp.java.tool-process-exec
- seceng.mcp.kotlin.tool-process-builder
- seceng.mcp.rust.tool-command-new
ASI03 Identity and Privilege Abuse
Section titled “ASI03 Identity and Privilege Abuse”355 rules · confidence P0 27, P1 197, P2 131 · severity error 109, warning 211, info 35
Edges inspected: reaches 38 · calls 44 · reads 25 · writes 16 · sends 22 · grants 188 · depends_on 5 · trusts 17
30 of 355 rules, surest first
- keryn.cryptoclaim.cs.jwt-verification-disabled
- keryn.cryptoclaim.go.jwt-verification-disabled
- keryn.cryptoclaim.java.jwt-verification-disabled
- keryn.cryptoclaim.kotlin.jwt-verification-disabled
- keryn.cryptoclaim.php.jwt-verification-disabled
- keryn.cryptoclaim.py.jwt-verification-disabled
- keryn.cryptoclaim.ts.jwt-verification-disabled
- keryn.skillcred.browser-credential-store-read
- keryn.skillcred.cloud-metadata-endpoint-access
- keryn.skillcred.container-escape-from-skill
- keryn.skillcred.macos-keychain-secret-read
- keryn.skillcred.os-credential-database-access
- keryn.skillcred.secret-manager-or-keyring-export
- keryn.skillcred.ssh-key-material-read
- seceng.agentcfg.mcp-docker-privileged
- seceng.agentcfg.mcp-docker-sensitive-mount
- seceng.agentcfg.mcp-env-inline-secret-known-prefix
- seceng.agentrules.secrets-to-url-command
- seceng.iacai.gcp-ai-role-granted-to-public
- seceng.mcp.py.dns-rebinding-protection-disabled
- seceng.mcp.py.oauth-client-secret-hardcoded
- seceng.mcp.py.token-verifier-unverified-jwt
- seceng.mcp.ts.dns-rebinding-protection-disabled
- seceng.mcp.ts.oauth-client-secret-hardcoded
- seceng.mcp.ts.verify-access-token-decode-only
- seceng.eval.public-dashboard-no-auth
- seceng.workflow.webhook-to-credentialed-action
- keryn.agents2026.claude-agent-sdk-bypass-permissions
- keryn.agents2026.claude-agent-sdk-ts-bypass-permissions
- keryn.browseragent.browser-launch-disable-web-security-js
ASI04 Agentic Supply Chain Vulnerabilities
Section titled “ASI04 Agentic Supply Chain Vulnerabilities”165 rules · confidence P0 23, P1 96, P2 46 · severity error 71, warning 83, info 11
Edges inspected: reaches 12 · calls 28 · reads 3 · writes 1 · sends 16 · grants 19 · depends_on 84 · trusts 2
30 of 165 rules, surest first
- keryn.agenthooks.copilot-hook-pipes-download-to-shell
- keryn.agenthooks.cursor-hook-pipes-download-to-shell
- keryn.agenthooks.gemini-hook-pipes-download-to-shell
- keryn.agenthooks.windsurf-hook-pipes-download-to-shell
- keryn.modelload.cloudpickle-or-dill-loads-py
- keryn.modelload.hf-hub-download-result-unpickled-py
- keryn.modelload.keras-load-model-safe-mode-disabled-py
- keryn.modelload.numpy-load-allow-pickle-py
- keryn.modelload.torch-hub-load-state-dict-nonliteral-url-py
- keryn.modelload.torch-hub-load-trust-repo-nonliteral-py
- keryn.modelload.yaml-load-unsafe-loader-py
- keryn.skillexec.reverse-shell-bash-devtcp
- keryn.skillexec.reverse-shell-netcat
- keryn.skillexec.reverse-shell-python
- keryn.skillexec.reverse-shell-script-langs
- keryn.skillexec.reverse-shell-socat
- seceng.agentcfg.hook-pipes-to-shell
- seceng.ciagent.agent-pr-head-checkout-on-pull-request-target
- seceng.llm05.taint.model-output-to-mcp-stdio-command-js
- seceng.mcp.py.stdio-server-command-from-request
- seceng.mcp.py.tool-description-hidden-unicode
- seceng.mcp.ts.stdio-client-transport-command-from-request
- seceng.mcp.ts.tool-description-hidden-unicode
- keryn.modelload.pandas-read-pickle-py
- keryn.modelload.request-path-to-model-loader-python
- keryn.modelload.torch-package-importer-py
- keryn.skillexec.decrypt-pipe-to-interpreter
- keryn.skillexec.download-chmod-run
- keryn.skillexec.download-then-sudo-run
- keryn.skillexec.download-to-temp-exec
ASI05 Unexpected Code Execution
Section titled “ASI05 Unexpected Code Execution”249 rules · confidence P0 65, P1 129, P2 55 · severity error 145, warning 94, info 10
Edges inspected: reaches 8 · calls 157 · reads 1 · writes 9 · sends 16 · grants 37 · depends_on 21
30 of 249 rules, surest first
- keryn.agenthooks.copilot-hook-pipes-download-to-shell
- keryn.agenthooks.copilot-hook-reverse-shell
- keryn.agenthooks.cursor-hook-pipes-download-to-shell
- keryn.agenthooks.cursor-hook-reverse-shell
- keryn.agenthooks.gemini-hook-pipes-download-to-shell
- keryn.agenthooks.gemini-hook-reverse-shell
- keryn.agenthooks.windsurf-hook-pipes-download-to-shell
- keryn.agenthooks.windsurf-hook-reverse-shell
- keryn.browseragent.model-output-to-page-script-js
- keryn.browseragent.model-output-to-page-script-py
- keryn.modelload.native-extension-library-load-py
- keryn.rustagent.model-output-to-process-command-rs
- keryn.rustagent.model-output-to-sql-rs
- keryn.rustagent.model-output-to-webview-eval-rs
- keryn.skillcred.container-escape-from-skill
- keryn.skillexec.reverse-shell-bash-devtcp
- keryn.skillexec.reverse-shell-netcat
- keryn.skillexec.reverse-shell-python
- keryn.skillexec.reverse-shell-script-langs
- keryn.skillexec.reverse-shell-socat
- keryn.skillpersist.crontab-write
- keryn.skillpersist.destructive-filesystem-command
- keryn.skillpersist.fork-bomb
- keryn.skillpersist.launchd-agent-install
- keryn.skillpersist.shell-rc-download-exec
- keryn.skillpersist.systemd-unit-enable
- keryn.skillpersist.windows-autorun-or-task
- seceng.agentcfg.hook-pipes-to-shell
- seceng.agentcfg.mcp-docker-privileged
- seceng.agentcfg.mcp-docker-sensitive-mount
ASI06 Memory and Context Poisoning
Section titled “ASI06 Memory and Context Poisoning”166 rules · confidence P0 2, P1 85, P2 79 · severity error 10, warning 117, info 39
Edges inspected: reaches 13 · calls 26 · reads 4 · writes 33 · sends 12 · grants 65 · depends_on 9 · trusts 4
30 of 166 rules, surest first
- keryn.rustagent.model-output-to-sql-rs
- seceng.eval.mlflow-log-raw-prompt
- seceng.agentrules.bidi-override-character
- seceng.agentrules.invisible-unicode-character
- seceng.asi06.langgraph-langchain-memory-write-from-untrusted-python
- seceng.asi06.memory-read-into-system-prompt-python
- seceng.asi06.memory-read-into-system-prompt-ts
- seceng.asi06.memory-shared-user-scope-python
- seceng.asi06.memory-shared-user-scope-ts
- seceng.ragloader.ingest-pipeline-fetch-user-url
- vulnfix-ai-ai_torch_load_unsafe
- keryn.browseragent.page-content-in-system-prompt-js
- keryn.browseragent.page-content-in-system-prompt-py
- keryn.outbound.model-output-to-crm-record-js
- keryn.outbound.model-output-to-crm-record-py
- seceng.agentrules.hidden-html-comment-instructions
- seceng.agentrules.instruction-override-phrase
- seceng.cloudflare-d1-ai-query-without-tenant-filter
- seceng.cloudflare-kv-ai-memory-without-tenant-prefix
- seceng.cloudflare-r2-public-ai-docs
- seceng.csharp-prompt-or-completion-logged
- seceng.deno.edge-kv-ai-memory-without-tenant-prefix
- seceng.eval.phoenix-trace-raw-data
- seceng.firebase-storage-public-ai-artifact
- seceng.firebase.firestore-ai-collection-read-without-owner-filter
- seceng.gateway.litellm-prompt-logging-enabled
- seceng.graphql-schema-exposes-prompt-or-tool-types
- seceng.java-prompt-or-completion-logged
- seceng.laravel-queue-ai-job-logs-prompt
- seceng.llm01.taint.retrieved-content-to-system-prompt-js
ASI07 Insecure Inter-Agent Communication
Section titled “ASI07 Insecure Inter-Agent Communication”57 rules · confidence P0 2, P1 35, P2 20 · severity error 17, warning 34, info 6
Edges inspected: reaches 14 · calls 9 · sends 6 · grants 17 · depends_on 1 · trusts 10
30 of 57 rules, surest first
- seceng.mcp.py.dns-rebinding-protection-disabled
- seceng.mcp.ts.dns-rebinding-protection-disabled
- keryn.trust.py.identity-assigned-from-model-output
- keryn.trust.ts.identity-assigned-from-model-output
- seceng.agentcfg.mcp-remote-plain-http
- seceng.asi07.a2a-push-notification-url-not-allowlisted-python
- seceng.asi07.a2a-push-notification-url-not-allowlisted-ts
- seceng.asi07.agent-event-stream-dispatches-commands-ts
- seceng.asi07.peer-agent-message-as-instructions-python
- seceng.asi07.peer-agent-message-as-instructions-ts
- seceng.mcp.csharp.token-passthrough
- seceng.mcp.go.token-passthrough
- seceng.mcp.py.token-passthrough
- seceng.mcp.ts.cors-wildcard-mcp-server
- seceng.mcp.ts.token-passthrough
- vulnfix-ai-ai_autogen_local_executor
- vulnfix-ai-ai_crewai_allow_code_exec
- keryn.mcp.py.client-plain-http-transport
- keryn.mcp.ts.client-plain-http-transport
- keryn.rustagent.llm-http-client-tls-verification-disabled-rs
- seceng.agentframework.crewai-delegation-with-dangerous-tools
- seceng.asi07.a2a-agent-card-without-security-schemes-python
- seceng.asi07.a2a-agent-card-without-security-schemes-ts
- seceng.llm01.taint.model-output-to-system-prompt-js
- seceng.mcp.go.go-sdk-http-bind-all
- seceng.mcp.go.mark3labs-base-url-cleartext
- seceng.mcp.go.mark3labs-http-bind-all
- seceng.mcp.py.network-transport-bind-all
- seceng.mcp.py.session-id-from-query
- seceng.mcp.py.uvicorn-mcp-app-bind-all
ASI08 Cascading Failures
Section titled “ASI08 Cascading Failures”84 rules · confidence P0 0, P1 39, P2 45 · severity error 9, warning 54, info 21
Edges inspected: reaches 7 · calls 12 · reads 4 · writes 6 · sends 4 · grants 48 · depends_on 1 · none 2
30 of 84 rules, surest first
- keryn.failopen.go.auth-error-ignored
- keryn.failopen.py.gate-exception-runs-action
- keryn.failopen.ts.gate-exception-runs-action
- seceng.asi08.agent-retry-without-cap-python
- seceng.asi08.agent-retry-without-cap-ts
- seceng.asi08.tool-error-swallowed-as-success-python
- seceng.asi08.tool-error-swallowed-as-success-ts
- seceng.asi08.unbounded-agent-fanout-from-model-output-python
- seceng.asi08.unbounded-agent-fanout-from-model-output-ts
- keryn.gaps.unbounded-model-loop
- keryn.vendor.anthropic-content-read-without-stop-reason-check-js
- keryn.vendor.anthropic-content-read-without-stop-reason-check-py
- keryn.vendor.openai-content-read-without-refusal-check-js
- keryn.vendor.openai-content-read-without-refusal-check-py
- seceng.cloudai.vertex-public-endpoint-or-logging
- seceng.llm05.taint.model-output-to-regex-js
- seceng.llm05.taint.model-output-to-regex-python
- seceng.llm10.js-ts-agentic-loop-dos-logic
- seceng.llm10.openai-call-without-max-tokens
- seceng.llm10.python-agentic-loop-dos-logic
- seceng.llm10.unbounded-retry-loop
- seceng.llm10.unbounded-retry-loop-js
- seceng.mongoose-ai-regex-query-from-model-output
- seceng.tsorm-ai-generated-orderby-or-select
- seceng.agentframework.crewai-process-without-rate-or-step-limit
- seceng.cloudai.azure-openai-public-network-access
- seceng.container-compose-ai-port-public-no-auth
- seceng.container.ai-service-loadbalancer-no-auth-hint
- seceng.deploy-ai-service-no-rate-limit-hint
- seceng.deploy.fly-ai-service-public-no-auth
ASI09 Human-Agent Trust Exploitation
Section titled “ASI09 Human-Agent Trust Exploitation”59 rules · confidence P0 3, P1 36, P2 20 · severity error 12, warning 31, info 16
Edges inspected: calls 15 · reads 4 · writes 1 · sends 1 · grants 26 · trusts 9 · none 3
30 of 59 rules, surest first
- keryn.outbound.model-output-as-payment-amount-js
- keryn.outbound.model-output-as-payment-amount-py
- seceng.llm09.missing-hallucination-disclaimer
- keryn.agents2026.claude-agent-sdk-bypass-permissions
- keryn.agents2026.claude-agent-sdk-ts-bypass-permissions
- keryn.agents2026.strands-bypass-tool-consent
- keryn.failopen.py.approval-error-path-returns-allow
- seceng.asi09.approval-callback-always-allows-python
- seceng.asi09.approval-callback-always-allows-ts
- seceng.asi09.approval-gate-disabled-python
- seceng.asi09.approval-gate-disabled-ts
- seceng.mcp.py.elicitation-auto-accept
- seceng.mcp.ts.elicitation-auto-accept
- keryn.agentclient.codex-approval-policy-never
- keryn.agentclient.gemini-auto-accept
- keryn.agentclient.vscode-chat-tools-auto-approve
- keryn.agents2026.claude-agent-sdk-can-use-tool-always-allows
- keryn.agents2026.vercelai-tool-needsapproval-disabled-destructive
- keryn.rustagent.model-output-to-desktop-open-rs
- seceng.agentrules.conceal-from-user
- seceng.llm05.taint.model-output-to-html-response-python
- seceng.llm05.taint.model-output-to-redirect-js
- seceng.llm05.taint.model-output-to-redirect-python
- seceng.llm05.taint.model-output-to-safe-markup-python
- seceng.llm06.langchain-tool-missing-args-schema
- seceng.schema-model-score-controls-security-decision
- seceng.dataset.synthetic-data-leaks-source-records
- seceng.forensics-model-version-missing
- seceng.forensics-no-user-tenant-on-ai-event
- seceng.forensics-prompt-hash-missing
ASI10 Rogue Agents
Section titled “ASI10 Rogue Agents”65 rules · confidence P0 14, P1 28, P2 23 · severity error 19, warning 33, info 13
Edges inspected: reaches 6 · calls 5 · reads 4 · writes 11 · sends 8 · grants 27 · depends_on 4
30 of 65 rules, surest first
- keryn.agenthooks.copilot-hook-reverse-shell
- keryn.agenthooks.cursor-hook-reverse-shell
- keryn.agenthooks.gemini-hook-reverse-shell
- keryn.agenthooks.windsurf-hook-reverse-shell
- keryn.skillpersist.crontab-write
- keryn.skillpersist.fork-bomb
- keryn.skillpersist.launchd-agent-install
- keryn.skillpersist.shell-rc-download-exec
- keryn.skillpersist.systemd-unit-enable
- keryn.skillpersist.windows-autorun-or-task
- seceng.asi10.agent-sandbox-host-escape-python
- seceng.asi10.agent-sandbox-host-escape-ts
- seceng.iacai.gcp-ai-role-granted-to-public
- seceng.eval.weave-trace-without-redaction
- keryn.skillpersist.dns-lookup-exfil-text
- keryn.skillpersist.reverse-tunnel
- seceng.iacai.azure-openai-network-acls-default-allow
- seceng.iacai.azure-openai-public-network-access-enabled
- seceng.iacai.k8s-ai-service-exposed-loadbalancer-nodeport
- vulnfix-ai-ai_autogen_local_executor
- keryn.agentcfg.hook-command-dns-lookup-with-variable
- keryn.skillpersist.disk-or-memory-fill
- keryn.skillpersist.git-hook-install
- seceng.cloudai.vertex-public-endpoint-or-logging
- seceng.container-ai-image-latest-or-unpinned
- seceng.gitlab-unpinned-ai-container-image
- seceng.iacai.azure-openai-no-network-acls
- seceng.iacai.vertex-ai-endpoint-no-private-network
- seceng.iacai.vllm-container-no-api-key
- seceng.mcp.v3.tool-file-read-from-args-py