How it works
Defenses
Attack Patterns MCP Attack Atlas What we catch Hardening manual OWASP LLM Top 10 MITRE ATLAS
Learn
Encyclopedia (new) Agent Security 101 Blog Reports CVP runs Thesis
Resources
Docs GitHub Action (live) vs Lakera vs Promptfoo Team
Theme

Changelog

Generated from the scanner changelog
0.6.11
Latest release
117
Releases listed
0.1.0
First release
Each card below names a release and links to its full entry, where the limits are stated.
Releases

117 releases

2026-09-26

0.6.0

8 sections, 63 pull requests
PENDING — open, not shipped, and not to be described as shipped
What this release claims about the proxy, and what it does not

Full entry and its limits

2026-08-12

0.4.0

3 sections
When this code actually reached you

Corrected: keyword countFrom detector to control

Full entry and its limits

2026-08-09

0.3.15

1 section
Day 17 — the V3 drain completes: cached credentials are not fresh approval, and evidence is not consent

Full entry and its limits

2026-08-08

0.3.14

1 section
Day 16 of the V3 drain — status output is not a signature: dry-runs, sync receipts, and auth introspection

Full entry and its limits

2026-08-07

0.3.13

2 sections
Day 15 of the V3 drain — repository instruction files, template comments, and generated receipts

Context

Full entry and its limits

2026-08-06

0.3.12

2 sections
Day 14 of the V3 drain — sampling parameters, speculative decode, and the carriers that hide in plain sight

Context

Full entry and its limits

2026-08-05

0.3.11

2 sections
Day 13 of the V3 drain — attestation lineage, credential-endpoint substitution, and provenance decay

Context

Full entry and its limits

2026-08-02

0.3.10

2 sections
Day 12 of the V3 drain — compaction artifacts, memory-state replay, and semantic-cache authority laundering

Context

Full entry and its limits

2026-08-01

0.3.9

1 section
Day 11 of the V3 drain — A2A agent cards, protocol-state envelopes, and delegation-bridge laundering

Full entry and its limits

2026-07-31

0.3.8

1 section
Day 10 of the V3 drain — phantom tool-result frames, browser accessibility metadata, and cache/transport evidence laundering

Full entry and its limits

2026-07-29

0.3.7

1 section
Day 9 of the V3 drain — the non-tool MCP surfaces, cached credential grants, and forged verdict receipts

Full entry and its limits

2026-07-28

0.3.6

1 section
Day 8 of the V3 drain — CI/CD and infrastructure tool output laundered into agent authority

Full entry and its limits

2026-07-21

0.3.5

2 sections
"A prompt injection is still a prompt injection through a tool" — tool_output channel-coverage fix
Day 7 of the V3 drain — tool-output responses claiming prior approval

Full entry and its limits

2026-07-12

0.3.1

2 sections
"Receipts" release — the mechanism layer + a reproducible benchmark

Context

Full entry and its limits

2026-06-27

0.2.69

3 sections
Fixed (false positives — credibility)

Added (17 net-new patterns)Internal

Full entry and its limits

2026-06-19

0.2.68

4 sections
Fixed (structured_metadata_poisoning detection — 16 patterns)
Fixed (false positives — credibility)

Added (10 net-new patterns)Internal

Full entry and its limits

2026-06-11

0.2.66

1 section
Added (discovery_file_poisoning expansion — agent config/discovery-file class)

Full entry and its limits

2026-06-07

0.2.63

2 sections
Added (V2 SHIP #9 — discovery_file_poisoning continued + repo_metadata_poisoning new category)

Context

Full entry and its limits

2026-06-06

0.2.62

2 sections
Added (V2 SHIP #8 — discovery_file_poisoning continued + FP credibility fix)

Context

Full entry and its limits

2026-06-04

0.2.60

1 section
Added (V2 SHIP #6 — cicd_metadata_poisoning + supply_chain_attestation_poisoning)

Full entry and its limits

2026-05-27

0.2.52

3 sections
Added (agent_workflow_security continuation + dogfood-failure fixes)
Fixed (May 27 dogfood-failure incident — first real use surfaced 4 bugs)

Context

Full entry and its limits

2026-05-26

0.2.51

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 7 + agentic runtime visibility blog)

Context

Full entry and its limits

2026-05-26

0.2.50

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 6)

Context

Full entry and its limits

2026-05-25

0.2.49

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 5)

Context

Full entry and its limits

2026-05-23

0.2.47

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 4)

Context

Full entry and its limits

2026-05-22

0.2.46

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 3)

Content

Full entry and its limits

2026-05-21

0.2.45

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 2)

Content

Full entry and its limits

2026-05-20

0.2.44

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 1)

Content

Full entry and its limits

2026-05-19

0.2.43

2 sections
Added (agent_workflow_security expansion — Cava research bundle chunk 8)

Content

Full entry and its limits

2026-05-17

0.2.42

2 sections
Added (cross_agent_injection + tool_output_poisoning expansion — Cava research bundle)

Content

Full entry and its limits

2026-05-16

0.2.41

2 sections
Added (tool_metadata_smuggling + tool_poisoning expansion — Cava research bundle)

Content

Full entry and its limits

2026-05-15

0.2.40

2 sections
Added (policy_scope_redefinition category expansion — Cava research bundle)

Content

Full entry and its limits

2026-05-15

0.2.39

4 sections
Added (Red-Team Gym hardening sweep — 3 cycles, 4 attack families closed to 0–1 misses)
Gym scoreboard (template adversary, 500 payloads/run)

TestsContext

Full entry and its limits