v1.5.3

What's Changed

• fix(llm): omit parallel_tool_calls on tool-less requests by @devin-ai-integration[bot] in #1029
• fix(container): reclaim abandoned browser sessions by @devin-ai-integration[bot] in #1034
• fix(container): keep the browser env alive where image ENV is dropped by @devin-ai-integration[bot] in #1035
• fix(container): write the browser profile as root by @devin-ai-integration[bot] in #1039
• fix(llm): send OpenRouter app attribution on the request itself by @devin-ai-integration[bot] in #1045
• chore: release v1.5.3 by @devin-ai-integration[bot] in #1046

Full Changelog: v1.5.2...v1.5.3

v1.5.2

What's Changed

• docs(skills): clearer description by @bearsyankees in #1013
• fix(tui): make the mount prompt clickable, and skip the mount instead of abandoning the scan by @0xallam in #1015
• test(tui): correct the nudge the internal-turn test asserts by @0xallam in #1016
• fix(agents): let an agent wait on what it already said by @0xallam in #1020
• fix(todo): stop a todo plan failing on priority or duplicates by @0xallam in #1021
• fix(sessions): open a sqlite connection per operation, not per thread by @0xallam in #1022
• fix(runner): resume after a user interrupt instead of failing by @0xallam in #1023
• fix(tools): coerce an empty-string list/dict argument to an empty container by @0xallam in #1024
• fix(runner): settle child agents before closing sessions at wind-down by @0xallam in #1025
• chore: release v1.5.2 by @devin-ai-integration[bot] in #1026

Full Changelog: v1.5.1...v1.5.2

v1.5.1

What's Changed

• expand firebase storage rules coverage by @bearsyankees in #1002
• fix(tui): scrollbar visibility, findings scrolling, and report navigation by @0xallam in #1006
• chore: release v1.5.1 by @devin-ai-integration[bot] in #1011

Full Changelog: v1.5.0...v1.5.1

v1.5.0

What's Changed

• fix(cli): align View label spacing in final panel by @devin-ai-integration[bot] in #918
• fix viewer tool call collisions across agents by @bearsyankees in #917
• fix: recoverable guardrail blocks and decoupled crash-notify by @devin-ai-integration[bot] in #919
• Scope viewer session cookie to the bound port by @yoni-at-strix in #922
• fix(cost): capture OpenRouter streamed usage.cost (fixes $0 kimi-k3 c… by @bearsyankees in #929
• fix(runtime): label docker sandbox containers with the run id for teardown by @devin-ai-integration[bot] in #933
• feat(llm): custom request headers for OpenAI-compatible endpoints via LLM_EXTRA_HEADERS by @devin-ai-integration[bot] in #936
• fix(llm): pass LLM_EXTRA_HEADERS through ModelSettings so they reach the agent loop by @devin-ai-integration[bot] in #937
• feat(llm): opt-in LLM_DISABLE_STREAMING for non-streaming OpenAI-compatible endpoints by @devin-ai-integration[bot] in #938
• fix(llm): prevent ChatGPT subscription scan freezes by @chunguscodes in #943
• fix(llm): surface structured provider refusals by @bearsyankees in #944
• fix: pre-v1-style lifecycle resilience — mailbox delivery, uniform revival, unexitable runner, waiting timeout, broader retries, crash-safe identity by @devin-ai-integration[bot] in #923
• docs(prompt): teach agents to recognize Caido proxy error pages instead of chasing them by @devin-ai-integration[bot] in #955
• fix(core): explicit lifecycle contract — plain text never ends a run; respond_to_user / wait_for_agents split by @devin-ai-integration[bot] in #954
• docs(llm-providers): explain the structured tool_calls requirement for local endpoints (#520) by @devin-ai-integration[bot] in #901
• feat(config): accept STRIX_REASONING_EFFORT=max for providers that support it by @devin-ai-integration[bot] in #956
• fix(tools): accept both the string and structured form of every tool argument by @devin-ai-integration[bot] in #957
• perf(cli): ~10x faster startup via lazy imports by @devin-ai-integration[bot] in #920
• fix(core): a child never leaves its parent waiting silently by @devin-ai-integration[bot] in #959
• feat(runtime): mount local targets instead of copying them in by @devin-ai-integration[bot] in #958
• fix: calibrate vulnerability severity to demonstrated impact by @bearsyankees in #962
• feat(tui): replace Textual with a Go/Bubble Tea interface by @kusonooyasumi in #941
• Create credential files with owner-only permissions by @lukiod in #945
• fix(tui): remove the shift+enter newline hint from the setup footer by @devin-ai-integration[bot] in #968
• feat: support API specs and Postman collections as targets by @5h4d0wr007 in #866
• fix(tui): keep a long error inside the status row by @0xallam in #970
• feat(reporting): record transitive dependency chain on SCA findings by @devin-ai-integration[bot] in #971
• fix(llm): keep tool-call ids unique so a recycled id can't erase history by @devin-ai-integration[bot] in #976
• fix(llm): cap the tool calls one assistant response may queue by @devin-ai-integration[bot] in #977
• fix(llm): abandon a model stream that stops producing events by @devin-ai-integration[bot] in #978
• fix(agents): collapse repeated waits queued inside one model turn by @devin-ai-integration[bot] in #979
• feat(reporting): structured reachability evidence ladder for dependen… by @bearsyankees in #974
• feat(reporting): require repo-relative manifest_path on dependency CVE findings by @devin-ai-integration[bot] in #982
• prompt changes by @devin-ai-integration[bot] in #984
• update readme by @devin-ai-integration[bot] in #985
• feat: agent-ready — installable SKILL.md skills, AGENTS.md, coding-ag… by @bearsyankees in #926
• docs(skills): require per-CVE affected-symbol matching in dependency reachability analysis by @devin-ai-integration[bot] in #986
• chore: release v1.5.0 + bump sandbox image to 1.3.0 by @devin-ai-integration[bot] in #987
• ci: fix Windows release build — TUI-sidecar archive check used forward-slash paths by @devin-ai-integration[bot] in #990
• ci: tolerate repr-escaped backslashes in the release TUI-sidecar check by @devin-ai-integration[bot] in #991
• ci: publish nested standalone archives as release assets by @devin-ai-integration[bot] in #994

New Contributors

• @chunguscodes made their first contribution in #943
• @lukiod made their first contribution in #945
• @5h4d0wr007 made their first contribution in #866

Full Changelog: v1.4.1...v1.5.0

v1.4.1

What's Changed

• fix(tls): replace raw urllib with requests for external HTTPS calls (frozen-build cert failures) by @yoni-at-strix in #903
• chore: release v1.4.1 by @devin-ai-integration[bot] in #912

Full Changelog: v1.4.0...v1.4.1

v1.4.0

What's Changed

• feat(cli): update notifications + self-update (strix --update) by @bearsyankees in #807
• chore(deps): bump cryptography to 49.0.0 and pyasn1 to 0.6.4 by @bearsyankees in #856
• fix(deps): cap cryptography <49 to keep Intel macOS universal2 wheel by @bearsyankees in #859
• Quit after scan instead of hosting local viewer by @yoni-at-strix in #858
• Pin release-workflow actions to commit SHAs + least-privilege token by @timhaines in #862
• chore(deps): bump setuptools to 83.0.0 in uv.lock by @0xallam in #872
• Sign in with a ChatGPT subscription for inference by @yoni-at-strix in #854
• docs(cli): document exit code 1 and clarify exit 0 semantics by @dvirarad in #871
• runtime: resolve staged local-dir path to avoid symlink rejection on macOS by @utkutugrul in #857
• feat(dedupe): add dedicated deduplication model by @kusonooyasumi in #823
• Rename root agent to Strix by @0xallam in #874
• fix(reports): strip markdown code fence from poc_script_code before rendering by @0xallam in #879
• refactor: move strix/viewer under strix/interface by @0xallam in #888
• docs(skills): remove references to tools not installed in the sandbox by @0xallam in #890
• feat(reporting): add read-only list_reports + get_report tools by @0xallam in #889
• feat(context): bound per-tool output before it enters agent history by @0xallam in #880
• feat(context): model-aware conversation compaction for long scans by @0xallam in #881
• feat(context): spill oversized tool output into the sandbox workspace by @0xallam in #882
• fix(runtime): retry transient mid-stream provider errors instead of crashing the scan by @0xallam in #891
• Add Linux ARM64 standalone release support by @MBemera in #886
• feat(llm): enable Bedrock/Anthropic prompt caching for Claude models by @seanturner83 in #772
• fix(cli): don't dump raw warm-up traceback over the LLM error panel by @0xallam in #896
• feat(runtime): graduated wrap-up warnings, budget reserve, and interactive budget pause/continue by @0xallam in #893
• fix(runtime): wake parent when a child hits a terminal state (MaxTurnsExceeded) by @0xallam in #897
• chore: release v1.4.0 by @0xallam in #898

New Contributors

• @timhaines made their first contribution in #862
• @dvirarad made their first contribution in #871
• @utkutugrul made their first contribution in #857
• @kusonooyasumi made their first contribution in #823
• @MBemera made their first contribution in #886

Full Changelog: v1.3.1...v1.4.0

v1.3.1

What's Changed

• sandbox: bump default inner-sandbox image to 1.1.0 by @devin-ai-integration[bot] in #850
• chore: release v1.3.1 by @devin-ai-integration[bot] in #853

Full Changelog: v1.3.0...v1.3.1

v1.3.0

What's Changed

• fix(container): keep /app/.venv/bin on the login-shell PATH so python… by @bearsyankees in #839
• fix(prompt): make root agent orchestrate-only and fold fixing into reporting by @devin-ai-integration[bot] in #827
• sandbox: shrink image 7.2GB → 3.8GB (cache cleanup, multi-stage Go build, drop ZAP) by @seanturner83 in #474
• Local viewer: UI polish and a Feedback & support tab by @0xallam in #847
• Local viewer: prominent scan switcher + rename to "pentest" terminology by @0xallam in #848
• chore: release v1.3.0 by @0xallam in #849

Full Changelog: v1.2.0...v1.3.0

v1.2.0

What's Changed

• fix(runtime): stage symlink-safe copies for LocalDir uploads by @bearsyankees in #766
• fix(runtime): retry transient sandbox startup failures by @bearsyankees in #768
• feat(runtime): resolve sandbox ports over a shared Docker network by @devin-ai-integration[bot] in #775
• fix(core): bound per-agent image memory (proactive budget + inherited-context scrub) by @devin-ai-integration[bot] in #779
• feat(runtime): opt-in resource limits for docker sandbox containers by @devin-ai-integration[bot] in #780
• Revert "fix(runtime): retry transient sandbox startup failures (#768)" by @devin-ai-integration[bot] in #782
• docs(prompts,skills): bound recon output for shared-disk hygiene by @devin-ai-integration[bot] in #783
• docs(tools): guide proportional wait_for_message timeouts by @devin-ai-integration[bot] in #784
• docs(prompts,skills): stop hardcoding /workspace/scratch path by @devin-ai-integration[bot] in #786
• fix(runtime): cap sandbox container logs to prevent host disk exhaustion by @0xallam in #785
• fix(runtime): close the docker client on session cleanup by @0xallam in #787
• fix(prompt): down-rate or skip findings on demo data / demo environments by @devin-ai-integration[bot] in #793
• fix(proxy,tooling): serialize+reconnect Caido client, actionable HTTPQL errors, sandbox tool guidance by @devin-ai-integration[bot] in #794
• revert(proxy): drop overfit Caido reconnect/HTTPQL band-aids, keep serialization lock by @devin-ai-integration[bot] in #799
• fix(report): restore cost tracking for OpenRouter and other LiteLLM-routed models by @devin-ai-integration[bot] in #801
• fix(llm): retry statusless mid-stream provider errors (quota/billing/transient) by @devin-ai-integration[bot] in #790
• fix(llm): add per-turn model request timeout so stalled streams fail fast and retry by @devin-ai-integration[bot] in #802
• fix(llm): use a JSON-serializable per-turn model timeout by @devin-ai-integration[bot] in #803
• recon asset discovery skill by @bearsyankees in #809
• docs(skills): add grafana_prometheus observability pivot skill by @bearsyankees in #812
• fix(proxy): recompute Content-Length when replaying a modified body by @thejesh23 in #816
• strip transfer encoding by @bearsyankees in #820
• fix(report): prevent code-fence breakout in vulnerability markdown by @thejesh23 in #817
• docs(reporting): add CVSS calibration guidance to reduce severity inf… by @bearsyankees in #821
• docs(skills): add Active Directory / Kerberos domain testing skill by @bearsyankees in #825
• Local run viewer: email reports, run history, and the platform suite by @yoni-at-strix in #813
• chore: release v1.2.0 by @devin-ai-integration[bot] in #826

New Contributors

• @thejesh23 made their first contribution in #816
• @yoni-at-strix made their first contribution in #813

Full Changelog: v1.1.0...v1.2.0

v1.1.0

Strix v1.1.0

First release since v1.0.4. Highlights below (~56 merged PRs).

New security skills

• Weak-password detection skill (#621, #654)
• LLM prompt-injection skill (#616)
• GCP and Auth0 reconnaissance skills
• Five new skills: OAuth, AWS, prototype pollution, deserialization, Django (#617)
• Skill-directory registration (#746)

SARIF / CI integration

• SARIF 2.1.0 emitter for GitHub code-scanning / ASPM ingestion (#626)
• STRIDE tagging of SARIF rules derived from CWE (#708)
• Omit SARIF version-control provenance for multi-repo scans (#726)

Dependency / SCA reporting

• New dependency reporting fields (#751)
• Require advisory_cvss for dependency findings + SCA TUI renderer (#753)

Cost & scope controls

• Configurable token / cost usage limits, --max-budget-usd (#576)
• Bind-mount option for large target repos (#577)
• --target list-file CLI option (#711)

LLM / provider

• Use core LiteLLM dependency (#752); cap openai<2.45 + litellm[proxy] so fresh installs run (#748)
• force_required_tool_choice setting (#730) and routed OpenAI required tool choice (#732)
• Scan-agent tool registration (#733); root scan prompt options (#750)
• Attribute OpenRouter usage to the Strix app (#760)
• Warn when the configured LLM is not a recommended/frontier model (#586)
• Route ollama models through ollama_chat so tool calling works (#562)
• Report cost for streamed OpenRouter calls (#634)

Reliability / fixes

• Default sandbox exec commands to Bash so source works (#764)
• Provider import-error hints for Bedrock/Vertex, full exception-chain walk (#588)
• Env vars win over persisted JSON across all aliases (#689)
• Graceful stop with resume hint on persistent RateLimitError (#593)
• Collapse child-agent initial input into a single user message (#589)
• Atomic CSV/MD writes to prevent corruption on crash (#628, #631); csv_path indentation (#637); avoid note-ID collisions (#630)
• Swallow torn-down docker socket on sandbox delete (#721); Caido HTTPS scheme (#722) + configured UI domains (#723)
• Keep verbose openai.agents DEBUG off sandbox stdout (#704)
• Strip ANSI escapes / control bytes from terminal tool output (#554)

TUI

• Reduce scroll stutter via throttled refresh + render caching (#687)
• "More content available" hint for long views (#687)
• Restore snappy sweep/progress animation frame rate (#759)

CI / telemetry / chores

• Lower Linux release glibc baseline (#707)
• Deduplicate scan-ended telemetry (#758); minor telemetry updates (#761); drop unhandled-exception collection (#585)
• Config-loader tests (#596); report-writer artifact tests (#667); uv.lock refresh (#606); README/docs updates

Contributors

Thanks to everyone who contributed to this release:

@0xallam @AtifAssari @Ayush7614 @Hardik-369 @Rome-1 @RudraDudhat2509 @Sonai124 @Stark-SK @ViperDroid @Zizouk22 @bastitva0-blip @bearsyankees @chirag127 @dpersek @e345ee @mhspektr @ousamabenyounes @sean-kim05 @seanturner83 @singe