mirror of
https://github.com/tiennm99/goclaw.git
synced 2026-10-15 06:16:10 +00:00
dev
15
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 ntduc
|
fe5a3c7f92 | fix(hooks): select prompt provider and model | ||
|
ntduc
|
03a196e613 | feat(hooks): add script context injection | ||
|
|
293aa2a5e1 |
(feat) MCP Server (#1365)
* feat(hooks/observe): add structured output validation hook (ObserveHook) - hooks/types.go: Add ObserveHook interface + BuiltInHookType enum - hooks/dispatcher.go: HookDispatcher emits ObserveHook with PhaseResult payload - pipeline/observe_stage.go: ObserveStage emits hook after ObserveResult built - pipeline/substates.go: ObserveStageResult carries hook results + validation errors - hooks/config.go: BuiltInHookTypeObserve added to BuiltInHookType enum PhaseResult carries structured output, token usage, tool calls + validation errors emitted post-ObserveStage. ObserveHook implementations can validate structured output against schemas, detect tool-call loops, enforce token budgets, etc. Hook fires after ObserveStage produces ObserveResult, before results propagate to next stage. ValidationError returned by hook halts pipeline and propagates error to caller without further stage execution. Co-Authored-By: Claude <noreply@anthropic.com> * ui(hooks): add post_model_response event to web UI - Add event to Zod schema, filter dropdown, and form dialog - Implement conditional test panel UI for model response payload - Add translations (en/zh/vi) for new test panel fields - Updated beta description to reference the new event Co-Authored-By: Claude <noreply@anthropic.com> * feat(mcp): add MCP CRUD server exposing goclaw resources at /api/mcp/ with Bearer token auth and X-GoClaw-Tenant-Id header, default to master tenant * feat(mcp): add goclaw_skills_write_file tool to edit skill files on disk The CRUD MCP server's goclaw_skills_update only touched skill DB metadata, with no way to edit a skill's SKILL.md/file content on the filesystem. Extract the versioned write logic from the web UI's skill file editor (SkillsHandler.handleWriteFile) into skills.WriteVersionedFile so both surfaces share identical validation and versioning, and expose it as a new MCP tool. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> --------- Co-authored-by: Bruno Clermont <bruno.clermont@gmail.com> Co-authored-by: Claude <noreply@anthropic.com> |
||
|
Bruno ClermontandBruno Clermont
|
cf80d90afb |
feat(hooks): add debug logging, capture console.log output, fail closed on matched-then-errored hooks (#1337)
prefilter() had zero logging -- matcher regex failures, tool-name mismatches, CEL compile errors, and CEL evaluating to false were all completely silent, making a misconfigured hook impossible to debug. A user's PreToolUse script hook silently failed to block an unauthorized file write with no way to determine why. - Added slog.Debug/Warn logging throughout prefilter() and the dispatch chain: hooks.prefilter.considered, matcher_checked, condition_evaluated, error, and hooks.dispatch.decision -- covering matcher match/miss, CEL compile/eval result, and final decision with reasoning. - console.log/console.error output from script hooks (captured via the existing goja sandbox stdout buffer, previously discarded after execution) now surfaces via slog.Debug and persists into HookExecution.ConsoleOutput / Metadata["console_output"] in the audit record, using the existing metadata JSON column (no migration needed). - Fail-closed, narrowly scoped: when a hook's matcher successfully matches a tool call but CEL condition or script execution then errors, the dispatcher now blocks that specific tool call (previously: silently treated as non-match, call proceeded allowed). Hooks that legitimately don't match remain unaffected pass-through -- this does not turn one broken hook into a global outage, it only blocks the specific gated action that couldn't be confidently evaluated. Added tests: matched-then-CEL-errors blocks the call and never invokes the handler; matcher-doesn't-match proceeds normally unaffected (confirms narrow scope); console.log output is captured in the audit record. Co-authored-by: Bruno Clermont <bruno.clermont@gmail.com> |
||
|
Goon
|
d64a31ebdb | fix(usage): enforce caps on auxiliary llm calls | ||
|
Duy Nguyen
|
5a189330f9 | fix(test): support windows test execution | ||
|
Duy /zuey/
|
4472c607b8 |
feat(workstation): Remote Workstation Runtime — SSH exec + security + audit (#4)
* feat(packages): add update flow for GitHub binaries (#900) Closes #900. Proactive update-check + atomic swap for GitHub-installed binaries on the Runtime & Packages page. Interfaces prepared for pip/npm/apk extension in Phase 2. - UpdateCache + UpdateRegistry + PackageLocker (ctx-aware keyed mutex) - GitHubUpdateChecker: ETag-aware, distinct /latest vs /list ETag keys, semver-correct ordering via golang.org/x/mod/semver, non-semver fallback that refuses to downgrade, pre-release + stable candidate fusion for the v1.0.0-rc.1 -> v1.0.0 transition - GitHubUpdateExecutor: two-phase .bak swap with hadBackup-aware rollback, manifest save retry (3x, 100ms/500ms/1s backoff), nil-safe meta access, explicit ScratchDir, 0755 set pre-rename - HTTP: GET /v1/packages/updates (SWR), POST /v1/packages/updates/refresh, POST /v1/packages/update, POST /v1/packages/updates/apply-all (always 200, failed[] is error source). Master-scope gated. - WS events package.update.{checked,started,succeeded,failed} forwarded to owner clients via event_filter.go - Frontend: useUpdates hook + 3 components (summary bar, update-all modal, row button), master-scope-gated disabled state - i18n: 8 backend keys + 17 frontend keys x en/vi/zh - Config: packages.github_token (reserved), updates_check_ttl, scratch_dir - 45+ new tests, race-clean, BenchmarkCheckAll10Packages ~1.1ms/op warm * docs(packages): document update flow + Phase 1 completion - packages-github.md: "Updating Installed Packages" section with UI + API contract, troubleshooting runbook (corrupt cache, rate-limit, scratch dir, mid-swap recovery) - 17-changelog.md + CHANGELOG.md: Phase 1 entry - 14-skills-runtime.md: cross-ref to update flow - journal entry capturing CRIT fixes (double-write, lock-key mismatch, rollback false-alarm) + design wins (keyed locks, red-team pre-flight) * feat(workstation): remote workstation runtime — SSH exec + security + audit Adds generic Remote Workstation Runtime enabling agents to execute commands on user-owned SSH workstations. Includes registry (DB + API + UI), SSH backend with connection pool and circuit breaker, workstation.exec + claude_remote tools, NFKC + binary-name allowlist security, and audit logging. Standard edition only. Closes #941. * fix(workstation): address 3 critical + 5 important code review findings - C1: Add json:"-" to Metadata/DefaultEnv fields; use SanitizedView() in all API responses to prevent SSH private key leakage - C2: Wire CheckEnv into PermCheckFn; LD_PRELOAD/PATH injection now blocked - C3: SSH Setenv fallback — prepend `export K=V;` when server rejects Setenv - I1: BackendCache sync.RWMutex → sync.Mutex (fix data race on lastUsed) - I2: Validate metadata shape in handleUpdate before store write - I3: Include command in exec-done event; activity sink uses actual cmd hash - I4: Wrap pool release in sync.Once (idempotent double-call safety) - I5: Verify workstation tenant ownership before adding permissions * fix(packages): bypass HTTPS+IP validation in update executor tests Test httptest servers bind to http://127.0.0.1 which fails both the HTTPS scheme check and literal-IP SSRF guard. Add testSkipDownloadValidation flag (same pattern as existing withTestDownloadHosts) to skip full URL validation in test context. * fix(workstation): address Claude review findings — tenant isolation + pool leak + dead code - Activity list: add workstation ownership check before listing (prevents cross-tenant activity enumeration via known UUID) - SSH pool: clean up p.sem + p.circuits maps in CloseWorkstation, prune, and Close to prevent unbounded map growth - RPC handlers: return ErrInvalidRequest on JSON unmarshal failure instead of silently using zero-value params - Remove unused containsControlChars function in normalize.go - HTTP tests: add 10s context timeout to prevent CI package timeout * fix(workstation): DefaultEnv JSON parse, backend cache leak, perm ownership check - DefaultEnv: replace KEY=VALUE text parse with json.Unmarshal (stored as JSON by HTTP handler, was silently ignored) - BackendCache: close losing backend on concurrent cache miss to prevent pruneLoop goroutine leak - Backend interface: add Close() error method; SSHBackend delegates to pool.Close() - handlePermList: add wsStore.GetByID ownership check (prevents cross-tenant UUID enumeration returning empty array vs 404) - scanRows: log scan errors instead of silently skipping * fix(workstation): wire activity sink shutdown + remove misleading comment - WireActivitySink: capture cleanup func, register in gateway shutdown (was discarded → retention goroutine leaked + buffered rows lost) - Add Stop() to WorkstationActivityStore interface (PG+SQLite already had it) - wireWorkstationTools returns cleanup func; gateway.go defers it - Remove misleading "re-validate env" comment in allowlist.go Check() * ci: bump unit test timeout from 90s to 120s hooks/handlers package (goja script tests) consumes ~85s on cold CI runners, leaving insufficient headroom for HTTP retry tests with 1s backoff. 120s provides adequate breathing room without masking real deadlocks. * fix: compile errors in integration tests + allowlist docstring - packages_update_test: add missing lockKey arg to registry.Apply - mcp_grant_revoke_test: remove unused fakeMCPClient struct - allowlist.go: fix Check() docstring to match actual 3-step pipeline * fix(test): relax mcp grant revoke assertion for pre-Phase02 state Execute-time grant checking not yet wired — test correctly gets an error but the message is "no active client" (nil clientPtr) rather than "grant revoked". Accept any error as valid regression guard. * chore: trigger CI on digitopvn/goclaw fork * ci: retrigger workflows * fix(permissions): classify workstation methods in RBAC policy |
||
|
viettranx
|
b676ee1351 |
refactor: adopt Go 1.26+ standard library modernizations
- Use maps.Copy in hooks dispatcher instead of manual map loop - Remove implicit loop variable capture in router_abort_test (Go 1.26 semantics) - Use range without index in http_test where index unused - Use min() builtin in script_test instead of manual min computation |
||
|
Kai (Tam Nhu) Tranandviettranx
|
4b02c27307 |
feat: native image_generation for Codex + OpenAI-compat providers (#1002)
* refactor(providers): migrate ToolDefinition.Function to pointer + add image response fields
ToolDefinition.Function becomes *ToolFunctionSchema with omitempty so native tool types (image_generation, web_search, etc.) can be declared without a function body. All 9 internal construction sites updated. CleanToolSchemas refactored — function-shape cleaning extracted into cleanFunctionSchema helper, outer pass-through handles native tools.
Added image response fields needed by the next commits: ChatResponse.Images, StreamChunk.Images, ImageContent.Partial (distinguishes partial frames from final images).
* feat(providers): native image_generation for Codex + OpenAI-compat tracks
Codex native (POST /codex/responses): emit image_generation tool object in request tools[] (type, action, model, output_format, partial_images). Handle SSE events response.image_generation_call.partial_image + response.output_item.done (type image_generation_call) + response.completed output[] walk for non-stream. Dedup per item_id. Extend codexSSEEvent/codexItem with output_format, result, partial_image_b64, partial_image_index.
OpenAI-compat (/v1/chat/completions): serialize ToolDefinition{Type:'image_generation'} as {type:'image_generation'} pass-through. Parse choices[0].message.images[] + delta.images[] (data URLs) via new parseDataURL helper; append to ChatResponse.Images.
ProviderCapabilities.ImageGeneration flag; Codex provider + adapter set true. Other providers default false.
* feat(agent,http,store): persist assistant images + tri-level image_generation gate
Agent loop tri-level gate: (provider capability) AND (AgentConfig.AllowImageGeneration, default true, stored in other_config.allow_image_generation) AND (request lacks x-goclaw-no-image-gen header). Gate in loop_tool_filter.go appends ToolDefinition{Type:'image_generation'} only when all three pass. Per-request opt-out parsed in chat_completions.go and propagated via RunRequest.NoImageGen.
Media persistence: persistAssistantImages writes final images (Partial:false) to {workspace}/media/{sha256}.{ext}, returns MediaRef entries, clears inline Images[] from the message. Idempotent on hash, traversal-safe, symlink-guarded. Invoked from pipeline.FinalizeStage via new Deps.PersistAssistantImages callback — covers both stream-final and non-stream paths.
Agent store reads AllowImageGeneration from other_config JSONB with absent/nil = true default (matches V3Flags pattern). No DB migration — code-only default.
* feat(ui/web): image_generation toggle + streaming placeholder + download filename
Composer chip 'Images' visible only when active agent's provider has ImageGeneration capability. Per-agent localStorage persistence via useImageGenToggle hook. When off, sends noImageGen:true to WS chat method (maps to x-goclaw-no-image-gen on upstream HTTP call path).
ActiveRunZone renders a skeleton placeholder while streaming partial_image frames arrive. MediaGallery assigns generated-YYYYMMDD-HHmmss.png as the download filename for hex/UUID PNGs.
i18n keys added to en/vi/zh chat.json: imageGenToggle, imageGenGenerating, imageGenDownloadName. 8 vitest tests for the toggle hook.
* docs: add Image Generation section to codebase-summary + changelog entry
Documents the new native image_generation pipeline across providers layer (Codex + OpenAI-compat), agent gate, media persistence, and web UI surface.
* fix(ui/web): match Codex-routed providers for image_generation toggle
Image-gen toggle visibility was hard-coded to provider id 'chatgpt_oauth' but real Codex-routed agents in production use provider ids like 'cliproxy-codex'. The toggle never rendered.
Replace the Set-has check with a small helper that accepts the literal ids plus any provider string containing 'codex' (case-insensitive). Same logic applied in both chat-input.tsx (composer chip) and chat-page.tsx (streaming placeholder gate).
Verified against a live Codex-routed agent: toggle now renders, noImageGen:true propagates on toggle-off.
* docs(pr-1002): targeted-mode UX evidence report
Captures the UI integration surface for native image_generation against a live Codex-routed agent on the remote dev backend.
Includes: composer toggle chip (rendered), streaming skeleton placeholder, and honest failure-path capture showing the legacy create_image builtin fallback. Self-contained HTML report in .github/pr-assets/1002/index.html.
* fix(permissions): classify sessions.compact as write method
CI RBAC-drift test (TestMethodRole_DriftCoverage_AllProtocolMethodsClassified) was failing because the new sessions.compact method added upstream was not classified in any of isPublicMethod / isAdminMethod / isWriteMethod / isReadMethod.
Sessions compaction mutates session history (compacts messages into summaries), so it belongs with the other sessions.* write methods.
* fix(tests): remove duplicate contains() in integration package
Both tts_gemini_live_test.go and mcp_grant_revoke_test.go declared a file-local func contains(s, substr string) bool with identical bodies, causing 'contains redeclared in this block' at compile time in the integration job.
Replace all call sites with strings.Contains (same semantics, stdlib) and drop the duplicates. No behavior change.
* feat(providers): NativeImageProvider interface + Codex implementation
Defines a provider-level contract (NativeImageProvider.GenerateImage) that OAuth-backed providers can implement to serve image generation without exposing static API credentials. Re-uses the PR's Track A native wire format (POST /codex/responses with image_generation tool, item.result decoding, SSE fallback).
CodexProvider + CodexAdapter implement it. Also adds MediaRef.Prompt field so downstream layers can propagate the generating prompt alongside the asset.
* feat(tools): route create_image through NativeImageProvider for OAuth providers
create_image.callProvider now checks for a NativeImageProvider implementation before the credentialProvider interface. When the provider chain points at a Codex-family provider (no static API key), the tool delegates to the provider's GenerateImage which executes the native ChatGPT Responses API image_generation flow.
Resolves 'provider X does not expose API credentials required for image generation' errors for openai-codex / cliproxy-codex chains. On success the tool embeds the user's prompt as a PNG tEXt 'Description' chunk (file-local pngEmbedPrompt helper to avoid tools→agent import cycle), writes the image to /tmp, and threads the prompt through result.MediaPrompts for downstream MediaRef propagation.
* feat(agent,pipeline): propagate image prompt through MediaRef + PNG tEXt embed helper
Adds EmbedPNGPrompt public helper in internal/agent/png_metadata.go that inserts a tEXt 'Description' chunk (plus 'Software: goclaw') into PNG byte streams before the IEND chunk. Non-PNG inputs are passed through unchanged — the helper never errors on unknown formats.
FinalizeStage wires MediaResult.Prompt (from create_image tool output) onto MediaRef.Prompt so the UI can render the generating prompt alongside the image. Per-image prompt list threaded via pipeline RunState.
* feat(ui/web): show generating prompt as caption under each image in MediaGallery
When a MediaRef carries a prompt, MediaGallery renders it as a muted italic caption (line-clamp-2) beneath the image with the full text in the title tooltip. Caption is hidden when the prompt is absent so non-assistant images (user uploads, legacy data) look unchanged.
MediaItem + session media_refs types extended with an optional prompt field; the chat-message adapter threads ref.prompt through when converting WS payloads to UI state.
* fix(providers/codex): stream:true + instructions for native image_generation
The ChatGPT Responses API on /codex/responses rejects two things hard:
- stream:false → HTTP 400 "Stream must be set to true"
- missing instructions → HTTP 400 "Instructions are required"
buildNativeImageRequestBody now sets stream:true and a purpose-specific instructions string ("Generate an image matching the user's description using the image_generation tool. Return only the image; do not describe it in text."). The existing parseNativeImageSSE path was already in place for stream parsing; routing changed from the non-stream branch to the SSE branch.
Regression assertions added to TestCodexGenerateImage_BuildsNativeRequest so these two fields can't silently regress.
* feat(providers,tools,ui): image_model whitelist (gpt-image-2 default, gpt-image-1.5 legacy)
Replaces the hardcoded "gpt-image-2" literal in buildNativeImageRequestBody with a user-configurable field threaded through NativeImageRequest.ImageModel. The whitelist is enforced by ValidateImageModel which rejects anything outside {gpt-image-2, gpt-image-1.5} with a clear error — prevents silent upstream 400s from model names the Responses API would reject.
create_image.callProvider reads params.image_model from the chain entry and threads it through. Empty / absent falls back to DefaultImageModel (gpt-image-2).
UI: added an 'Image model' select inside the existing openai-codex Settings panel on the Create Image Provider Chain dialog. Options: Default · gpt-image-2 (recommended) and Legacy · gpt-image-1.5. i18n keys in en/vi/zh tools.json under builtin.mediaChain.
Tests: TestCodexGenerateImage covers default/legacy/rejected model cases; TestCreateImageTool_ThreadsImageModel covers params→request threading with empty/legacy/explicit sub-cases.
* fix(tools): raise media chain default timeout to 600s/1 retry for image gen
gpt-image-2 on complex prompts (dense Vietnamese text, infographic layouts) legitimately takes 4–8 minutes to complete. The previous default of 120s × 2 retries routinely died mid-generation with 'context deadline exceeded' — the upstream run was still producing bytes when our ctx cancelled.
Default is now Timeout: 600 / MaxRetries: 1. Retries dropped to 1 because image generation is stateful per upstream run: a mid-flight timeout leaves orphan server work, and retrying a fresh generation doubles cost for no gain. Surface the failure fast so operators can widen the timeout instead.
Operators can still set a tighter value explicitly via the Chain dialog.
* refactor: remove user-facing Images toggle, keep admin-level AllowImageGeneration
The per-request opt-out toggle (composer chip + streaming placeholder + noImageGen header plumbing) was a support footgun — users toggle OFF, forget, then can't generate images and think it's broken. Removed in full.
Kept: AgentConfig.AllowImageGeneration (admin kill-switch, stored in other_config.allow_image_generation, default true). Tri-level gate simplifies to two tiers: provider capability AND agent config allows.
Removed: useImageGenToggle hook, IMAGE_GEN_PROVIDER_IDS set in chat-input, supportsImageGenProvider helper, agentProvider/agentKey props on ChatInput, showImageGenPlaceholder prop on MessageBubble/ActiveRunZone/ChatThread, noImageGen param on use-chat-send, parseNoImageGen in chat_completions.go, NoImageGen on RunRequest, no_image_gen_header_test.go, imageGenToggle/imageGenGenerating i18n keys. Kept imageGenDownloadName — used by MediaGallery for generated-\*.png filename resolution.
* docs(pr-1002): refreshed UX trace + updated codebase notes
Replaces the earlier stealth-state evidence with a clean three-capture trace from a real successful run: inline image + prompt caption, MediaGallery lightbox expansion, Chain dialog with the new Image model dropdown open. Skill-routing rows scrubbed from the capture — they reflect per-agent skill setup, not anything this PR introduces.
codebase-summary + changelog: reflect final state (toggle removed, image_model selector, 600s default chain timeout, gpt-image-2 as quality baseline).
* fix(pipeline): preserve mid-loop image_generation output across iterations
FinalizeStage previously read state.Think.LastResponse.Images, which holds
only the last iteration's response. If the LLM emitted image_generation_call
in iteration N alongside a function_call, then responded text-only in N+1,
the image from N was silently dropped on finalize.
Accumulate final (non-partial) images into state.Observe.AssistantImages
across every iteration via ObserveStage, and source FinalizeStage from the
accumulator instead of LastResponse. Partial streaming frames are filtered
defensively; response.Images is cleared on drain to prevent double-counting
on re-exec.
* test(pipeline): regression coverage for mid-loop image accumulation
Six ObserveStage cases covering image accumulation semantics:
- single-iter image-only, image+tool_call same iter, mid-loop image
surviving text-only final iter, multiple images across iters, partial
frame filtering, nil-response safety.
Two FinalizeStage cases verifying accumulator is the source of truth:
- PersistsFromObserveAccumulator: image in Observe + empty LastResponse
must still be persisted via PersistAssistantImages.
- NoPersistWhenAccumulatorEmpty: no call when no images were emitted.
---------
Co-authored-by: viettranx <viettranx@gmail.com>
|
||
|
viettranx
|
f3d6c99dda |
fix(hooks): close source-tier forge + i18n + concurrency hardening
Code-review findings from Wave 1 audit (commits 097776c5..f76a681e).
Critical (security):
- C1 internal/gateway/methods/hooks.go: parseHookConfigParams strips
caller-supplied source/id/created_by/version. Without this a tenant
admin could POST {"source":"builtin"} and escalate their UI hook into
the dispatcher's builtin capability tier (which is allowed to mutate
event input). Default Source resolves to "ui" via Validate.
- C2 internal/gateway/methods/hooks.go: handleUpdate also strips source +
created_by from the patch map. Same forge surface via PATCH instead of
POST.
High:
- H1 internal/gateway/methods/hooks.go: Update + Delete handlers wrap
hooks.ErrBuiltinReadOnly into the i18n key MsgHookBuiltinReadOnly so
users get a localized "builtin hooks are read-only" message instead of
the raw English sentinel string.
- H3 internal/hooks/handlers/script.go: bound tenantSems map with
opportunistic sweep — once map crosses 64 entries, idle slots
(no in-flight grants AND last-used > 1h ago) are reclaimed under the
same lock the acquire path already takes. Goroutine-free; common-path
cost unchanged.
Medium:
- M1 internal/hooks/dispatcher.go: builtinAllowlistLookup uses
atomic.Pointer instead of plain package var. Fixes the parallel-test
race where one test installs a lookup while another reads.
- M3 ui/web/index.html: drop http: https: from CSP connect-src — keep
'self' + ws:/wss: only. Same-origin XHR/WS still works; cross-origin
HTTP is no longer wildcard-permitted.
Tests:
- internal/gateway/methods/hooks_source_strip_test.go pins the C1+C2
strip behavior at the parse layer with a regression case that posts
every forge field and asserts each is zeroed.
|
||
|
viettranx
|
a6592fd65f |
feat(hooks/builtin): ship pii-redactor exemplar — email + phone masking
First real builtin under Phase 04 infrastructure. Registers two DB rows
(user_prompt_submit + pre_tool_use) on startup, runs the embedded JS with
source-tier gate + dispatcher dotted-path allowlist.
- internal/hooks/builtin/pii-redactor.js (ES5.1, goja-compatible):
handle(event) returns updatedInput only when email or E.164-ish phone
actually matched. Redacts rawInput + toolInput.{command,query,content}.
- builtins.yaml: pii-redactor entry (version 1, priority 900, timeout 2s,
on_timeout=allow — missing redaction < blocking conversation).
- Drop _placeholder.js; pii-redactor.js is now the first real //go:embed match.
- Export builtin.Source(name) so handler tests load embedded JS without
going through Load()/cache state.
- Tests:
* loader_test: registry contains pii-redactor after Load().
* script_pii_redactor_test (handler-level, real goja runtime): email
masked in rawInput; +E164 phone masked; toolInput.command redacted,
toolInput.path left alone (not in mutable_fields); no PII → no
mutation; redact(redact(s)) == redact(s).
* dispatcher_test: 3 dotted-path walker cases (toolInput.command only
lets that key through; rawInput-only doesn't let forged toolInput
override original event keys; toolInput wildcard merges all keys).
Bench: ~242µs/op on 1 KiB Apple M4 Max, well under <1ms target.
|
||
|
viettranx
|
a556161d0d |
feat(hooks/script): goja ES5.1 script handler + deny-all sandbox
Wave 1 Phase 02 — pure runtime + sandbox. Dispatcher wiring, edition gate,
and DB migration land in Phase 03.
Handler:
- internal/hooks/handlers/script.go — ScriptHandler with two-layer semaphore
(global 10 / per-tenant 3), hashicorp/golang-lru v2 program cache (cap 500),
InvalidateHook hook for Phase 03, watchdog goroutine interrupting on ctx
cancel, sanitized error text (strips \n + goja " at <frame>" inline tails)
- internal/hooks/handlers/script_sandbox.go — deny-all/allowlist hardening,
SetMaxCallStackSize(256), prototype-chain nullify running BEFORE deny pass
so Function.prototype.constructor can still be reached
- internal/hooks/handlers/script_runtime.go — LRU compile, JSON round-trip +
deep-freeze event binding, 4KiB truncating stdout capture, strict return
parser enforcing {decision, reason?, updatedInput?}
Types:
- HandlerScript HandlerType = "script"
- DecisionAsk / DecisionDefer reserved (Wave 1 treats as block + warn)
- hooks.ScriptResult + WithScriptResult / ScriptResultFrom ctx helpers for
Phase 03 dispatcher to apply builtin-source UpdatedInput
Tests (go test -race green):
- script_test.go — 12 unit cases (happy/error/timeout/ask/defer/truncate/H1)
- script_sandbox_corpus_test.go — 27 cases: escape primitives (constructor
chain, Reflect, Proxy, Symbol, Promise, __proto__ walk, eval, Function,
Date chain, JSON replacer, toJSON DoS), resource bombs (recursion, loop,
memory, ReDoS), typeof sanity (Reflect/Proxy/Symbol/Promise/GoError/
globalThis/eval/Function undefined), mutation defense (toolInput/rawInput
frozen, Go-side map unchanged)
Deps: github.com/dop251/goja, github.com/hashicorp/golang-lru/v2.
Gates: go vet clean, go test -race clean, go build ./... + sqliteonly clean.
|
||
|
viettranx
|
553340b3a9 |
feat(hooks/prompt): LLM-driven prompt handler with per-turn cap
Adds PromptHandler that runs an LLM over matched hook events (user prompts, tool calls) to produce decisions (allow/block/modify). Features: - Budget integration: per-tenant token spend enforced via budget.Store - Per-turn counter seeded in ContextStage and propagated through state.Ctx so PreToolUse fires under the same cap as UserPromptSubmit - prompt_template required (non-empty) at validate time to prevent misconfigured hooks from silently no-oping - Injection-hardened: tool_input/raw_input are quoted into system prompt rather than concatenated - Resolver abstraction (RegistryResolver) picks model from provider registry with SystemConfigs fallback |
||
|
viettranx
|
fb3552f7e5 |
fix(hooks/security): add SSRF-safe http client missed by phase 2 commit
Phase 2 (a587231b) advertised SSRF hardening for the http hook handler but the supporting `internal/security` package was never created, so the production HTTPHandler fell back to a bare http.Client and admin-config webhooks could probe loopback / link-local / RFC1918 / cloud-metadata. - internal/security/ssrf.go: Validate(rawURL) parses + resolves once, rejects loopback/link-local/private/multicast/unspecified + 169.254.169.254; NewSafeClient(timeout) returns an http.Client whose DialContext pins the resolved IP from context (defense-in-depth re-checks the dialed IP) and refuses redirects (CheckRedirect = ErrUseLastResponse) - internal/hooks/handlers/http_handler.go: call Validate before each request, attach pinned IP via security.WithPinnedIP(ctx, ip) - cmd/gateway_managed.go: construct HTTPHandler with Client: security.NewSafeClient(10*time.Second) - tests: 13 new ssrf_test.go cases (every block category + redirect refused + dial pinned); existing http_test.go retrofitted with security.SetAllowLoopbackForTest helper for httptest.NewServer - plan: phase-02 Step 2a marked done with reference to this commit Refs: GitHub Issue #875 |
||
|
viettranx
|
97f9784443 |
feat(hooks): phase 2 — handlers, pipeline wiring, ssrf-safe http client
- add command + http handlers (internal/hooks/handlers) with edition gating and Authorization-header decryption via crypto.Decrypt - add SSRF-safe dialer (internal/security/ssrf.go): DNS-once + pinned IP, blocks loopback / link-local / private ranges - wire dispatcher into pipeline stages: ContextStage fires SessionStart (async) + UserPromptSubmit (sync); ToolStage fires PreToolUse (sync, COW staging for updatedInput) + PostToolUse (async); FinalizeStage fires Stop (async) - bridge delegate events to dispatcher (SubagentStart / SubagentStop) - construct dispatcher in cmd/gateway_managed.go with both handlers and thread it through agent.ResolverDeps + delegateTool.SetHookDispatcher - unit tests for both handlers + integration tests covering HTTP allow with audit row, HTTP block, command Lite-only gate, delegate bridge subscribe - defer worker pool optimization (Phase 2 Step 3) to Phase 4 Refs: GitHub Issue #875 |