#311 let callers point GET /osm/api/registry-info at any registry via the
registry_url query param. Both modes then called IsBinaryInstalled() on every
entry, which runs `sh -c <valide-command>` — so a GET with a hostile registry
executed arbitrary shell on the server. With SameSite=Lax session cookies and
reflect-all CORS, that was reachable by CSRF from any page an operator visits.
- add installer.IsBinaryInstalledNoExec() and use it whenever registry_url is
set; only the embedded registry is trusted to run validate commands
- match isGitHubURL() on the parsed hostname so a lookalike host such as
evil.tld/?x=github.com no longer receives the GitHub token
- cap remote registry reads at 32MB instead of an unbounded io.ReadAll
- surface LoadRegistry errors in nix-build mode rather than returning a
success response with all metadata silently missing
- report the same registry_url semantics in both modes, and document the
no-exec behaviour in docs/api/install.mdx
Vigolium http_record and finding entries can embed multi-MB response bodies,
producing JSONL lines that exceed bufio.Scanner's max token size and fail
with 'token too long'. Switch to bufio.Reader.ReadString, which grows as
needed, and extract per-line logic into a processLine closure. Add a
regression test that imports a finding with a 12MB response body.
- Emit the raw `threads` value in step command summaries so template variables
like `{{deparosParallel}}` survive until templateEngine.Render resolves them,
instead of silently collapsing to 1 when Int() fails to parse the template
- Only print the Artifacts section in `printResultSummary` when -v/--verbose
is set, removing state/log file clutter from default run output
- Update embedded UI build artifacts under public/ui/_next to the latest
dashboard build (chunk renames, manifest updates, page txt/html refreshes)
- Reorder Migrate() to add columns to existing tables BEFORE creating
indexes, since CREATE TABLE IF NOT EXISTS is a no-op on existing
databases and some indexes (e.g. finding_hash) reference migrated
columns that would otherwise be missing
- Add explanatory comment in internal/database/database.go documenting
why column migrations must precede index creation
- Surface a helpful CLI hint in pkg/cli/root.go pointing users to
'osmedeus db migrate' when startup fails with 'failed to run
migrations', so stale schemas are easier to recover from
- Add db_import_vigolium JS function that routes vigolium JSONL records
by envelope type: http_record -> assets, finding -> vulnerabilities
(deduped on new finding_hash column), skipping scan/oast_interaction
- Add FindingHash field to Vulnerability model with idempotent ALTER
TABLE migration and matching index
- Suppress run errors when using --silent --empty-target placeholder mode
- Retry transient fetchURLContent failures (network errors, 408/429/5xx)
with exponential backoff; leave 4xx untouched so GitHub auth fallback
can engage
- Add unit tests for the vigolium importer (import + idempotency) and
the HTTP retry behavior, plus a vigolium juice-shop sample fixture
- Bump katana, naabu, kingfisher, bearer registry entries and add
vigolium to the direct-fetch registry
- Track tmux sessions in ActiveRun and kill them on run cancel via new
TmuxHooks indirection; expose killed_tmux_sessions in CLI and API responses
- Add ExecuteSSHCommand with remote pidfile + process-group kill watcher so
cancelling a run actually terminates remote scans (not just the local session)
- Route ssh_exec/ssh_rsync/sync_* through the run's cancellable context via
new RunContextHooks
- Switch docker-publish to sequential per-arch buildx builds + imagetools
manifest to avoid OOM on multi-arch builds; add docker-buildx-setup target
- Cross-compile Dockerfile via BUILDPLATFORM/TARGETOS/TARGETARCH and retry
SAST binary installs to survive QEMU-flaky downloads
- Bump version to v5.0.3
Workers now propagate RunUUID from the submitted task so the master
can resolve the correct DB run record and associate step results.
Also increments completed_steps on successful insert and connects
the worker to the database for db_import_* functions.
Build was being killed by the runner ~29min in (OOM on 7GB ubuntu-latest)
because goreleaser builds all targets in parallel by default. Cap to 2,
drop linux/386, and tune GOGC for lower peak memory.
- Upgrade goreleaser/goreleaser-action from v6 to v7 in manual and nightly release workflows
- Add timeout-minutes and Go cache to nightly release for faster, safer builds
- Normalize YAML string quoting from single to double quotes
- Add 'optional' tag to jq and massdns in direct-fetch registry metadata
- Auto-install Pulumi CLI when not found instead of returning error
- Add installed binary to PATH for immediate availability
- Update SSH pool retry config to 5 attempts with 10s fixed delays
- Improve user feedback with progress indicators
- Update subfinder from 2.12.0 to 2.13.0 with corresponding download URLs
- Update nuclei from 3.7.0 to 3.7.1 with updated release links
- Update httpx from 1.8.1 to 1.9.0 across all platforms
- Update katana from 1.4.0 to 1.5.0 with new download links
- Update naabu from 2.4.0 to 2.5.0 with refreshed URLs
- Update kingfisher from 1.76.0 to 1.88.0 with new release binaries
- Update bearer from 2.0.0 to 2.0.1 with updated links
- Update gosec from 2.22.11 to 2.24.7
- Update interactsh from 1.3.0 to 1.3.1
- Update cloudfox from 1.17.0 to 2.0.1
- Update last_update_at timestamp to 2026-03-13T15:11:36.292828+00:00
- Extract KillProcessAndChildren into core.types for reuse across CLI and server handlers
- Replace hardcoded 'claude-code' strings with core.DefaultACPAgent constant
- Precompute absolute allowed paths in ACP client to avoid repeated filepath.Abs calls
- Simplify runQuerySteps to delegate to runQueryTable, reducing code duplication
- Refactor agent_chat concurrency guard using sync.Mutex.TryLock for cleaner code
- Use request context for agent timeout instead of background context
- Add query subcommands (vulns, runs, steps) with JSON output and flexible filtering
- Enhance assets command with --where, --search, --value filters (fuzzy matching)
- Expand asset searchable columns (status_code, content_type, title, dns_records, tls, tech)
- Add run status and run cancel subcommands with process termination
- Support control-plane and PID-based cancellation with database updates
- Implement fuzzyFilters in GetTableRecords for case-insensitive substring matching
- Add OSM_IGNORE_REGISTRY env var to skip auto binary installation and health checks
- Update health check to display warning when binary folder is empty but OSM_IGNORE_REGISTRY is set
- Update installRequiredBinaries to skip installation when OSM_IGNORE_REGISTRY=true
- Enhance DownloadFile with retry logic and wget/curl fallback for resilient downloads
- Refactor download function to validate content length and handle transient failures
- Refactor test code to use switch statements instead of if-else chains
- Implement agent-acp step type for spawning external ACP agent subprocesses via Agent Communication Protocol
- Add ACPExecutor with validation, field rendering, and subprocess lifecycle management
- Integrate agent-acp field rendering in StepDispatcher (batch and sequential modes)
- Add run_agent() utility function for workflows to execute ACP agents from steps and JS context
- Add osmedeus agent CLI command for interactive agent execution with --agent, --cwd, --timeout, --stdin, and --list flags
- Add /osm/api/agent/chat/completions REST endpoint with OpenAI-compatible chat format and concurrency control
- Support agent selection via: built-in names (claude-code, codex, opencode, gemini) or custom acp_config.command
- Add step-level configuration: cwd, allowed_paths, acp_config (command, args, env, write_enabled)
- Add comprehensive E2E tests for agent-acp workflows (basic, minimal, config, codex variants)
- Add test workflows in test/testdata/workflows/agent-and-llm/
- Update AGENTS.md documentation with agent-acp examples, CLI usage, and API endpoints
- Add tracking of trigger input variables and recognize their availability at runtime
- Downgrade undefined variable warnings to info-level when variables come from event trigger inputs (both new and legacy syntax)
- Add event envelope variables (EventTopic, EventSource, EventTimestamp, etc.) as recognized built-ins when workflow has event triggers
- Include comprehensive test coverage for trigger input scenarios, event triggers, and mixed variable definitions
- Add platform variables (PlatformOS, PlatformArch, PlatformInDocker, PlatformInKubernetes, PlatformCloudProvider) to linter's built-in variables list
- Extract checkStepFieldsForUndefinedVars and collectReferencedVarsFromStep helper functions to support comprehensive scanning of all template-renderable fields including parallel_functions, speed_args, config_args, agent fields, HTTP headers, LLM messages, decision cases/conditions, and memory paths
- Add recursion for nested steps (foreach inner steps and parallel_steps) with proper variable scoping and support for loop-specific variables (_id_, foreach variable)
- Enhance message formatting with highlightQuoted() to make quoted terms bold+yellow in error output
- Enable UndefinedVariableRule by default in GetDefaultRules()
- Add comprehensive test coverage for newly supported fields and nested step types (TestUndefinedVariableRule_*, TestUnusedVariableRule_ReferencedInNewFields)
- Add test fixture and E2E test (TestWorkflow_Validate_UndefinedVarsComprehensive) validating detection across functions and foreach
j3ssie