4f8e16f5cd
feat: implement S7/S8/S9 SkyEye thought-logic review + evolution + full-member identity
...
S7: Mandatory pre-execution SkyEye review for ALL instructions (cannot be disabled)
- 4 review dimensions: decision pattern, logical coherence, expression, intent
- 3 outcomes: pass / suspect (suspend) / reject
- Registered as global middleware in server.js
S8: SkyEye evolution - auto-accumulates from all agent runtime logic
- Per-person behavior history (200 records each)
- Pattern summarization: methods, paths, active hours
S9: Full-member thought-logic identity verification
- Per-person identity maturity model with cold-start threshold (10 actions)
- Cold-start members auto-suspended for manual confirmation
- Self-reinforcing: more participation = stronger identity model
Config: skyeye-policy.json v2.0.0, server.js v5.0.0
Co-authored-by: qinfendebingshuo <207279273+qinfendebingshuo@users.noreply.github.com>
Agent-Logs-Url: https://github.com/qinfendebingshuo/guanghulab/sessions/6ee737eb-684b-44b8-aeaa-f73eb7c6f9c2
2026-03-25 09:18:52 +00:00
0a63716226
feat: implement S6 personal channel full autonomy rule
...
S6 Rule: Developers have full autonomy within their own channels.
- Add isChannelSelfDeploy() in autonomy-engine.js (single source of truth)
- Add channel self-deploy bypass in approval.js POST /request
- Add channelAutonomyRule S6 config in autonomy-rules.json
- Own-channel deploys skip SkyEye/approver authorization
- Cross-channel/system changes still go through full S2 approval
Co-authored-by: qinfendebingshuo <207279273+qinfendebingshuo@users.noreply.github.com>
Agent-Logs-Url: https://github.com/qinfendebingshuo/guanghulab/sessions/6ee737eb-684b-44b8-aeaa-f73eb7c6f9c2
2026-03-25 09:04:51 +00:00
e0880deb46
feat: implement Phase 7 Execution Protection Layer (ZY-LANGDRIVE-S1)
...
L1 Frontend Isolation:
- docs/js/execution-guard.js: UI lock/unlock, execution replay polling, browser guard
- docs/css/execution-guard.css: Execution lock styles, progress bar, log entries
L2 API Execution Interception:
- backend/api-server/middleware/execution-lock.js: Execution context lock, cancel interception
- backend/api-server/routes/execution.js: Status query, cancel/delete rejection (403)
L3 Execution Atomicity:
- backend/api-server/services/checkpoint.js: Checkpoint system with rollback
- backend/api-server/services/atomic-executor.js: Atomic execution wrapper
- backend/api-server/services/execution-watchdog.js: 15min timeout, deadlock prevention
Integration:
- Updated server.js to v3.0.0: execution routes, cancellation middleware, watchdog startup
Co-authored-by: qinfendebingshuo <207279273+qinfendebingshuo@users.noreply.github.com>
Agent-Logs-Url: https://github.com/qinfendebingshuo/guanghulab/sessions/6ee737eb-684b-44b8-aeaa-f73eb7c6f9c2
2026-03-25 08:37:56 +00:00
5e8bb969a4
fix: address code review feedback for LANGDRIVE implementation
...
- Normalize module paths in deploy permission checks (trailing slash)
- Add devId format validation in PATCH /dev/:devId/status
- Fix onboarding resume logic (allow resuming incomplete sessions)
- Switch audit logging from sync to async file I/O
- Add hasModuleAccess helper for consistent module permission checks
Co-authored-by: qinfendebingshuo <207279273+qinfendebingshuo@users.noreply.github.com>
Agent-Logs-Url: https://github.com/qinfendebingshuo/guanghulab/sessions/a260ce18-279a-4490-af90-a90a7b1f46cd
2026-03-25 08:27:00 +00:00
38d8afef55
feat: implement language-driven OS capabilities (ZY-LANGDRIVE)
...
Phase 3 - Write Capability Layer:
- Add backend/api-server/routes/write.js with 10 write endpoints
- Extend services/github.js with triggerWorkflow + createFile
- Add middleware/audit.js for operation audit logging
- Rate limiting: 10 writes/min per developer
Phase 4 - Intent Router Engine:
- Add config/function-tools.json with 10 operation definitions
- Add middleware/intent-router.js for NL → API routing
- Add docs/js/intent-tools.js for frontend function calling
Phase 5 - Permission Sandbox System:
- Add config/permissions.js with 4-level model (L0-L3)
- Add middleware/auth.js (identity + permission check)
- Add middleware/sandbox.js (preview environment guard)
Phase 6 - Cognitive Guidance System:
- Add config/onboarding-script.json (5-round guided dialogue)
- Add routes/onboarding.js (onboarding API + tools endpoint)
- Add docs/js/onboarding.js + docs/css/onboarding.css (UI)
Update server.js to v2.0.0 with all new routes and middleware
Co-authored-by: qinfendebingshuo <207279273+qinfendebingshuo@users.noreply.github.com>
Agent-Logs-Url: https://github.com/qinfendebingshuo/guanghulab/sessions/a260ce18-279a-4490-af90-a90a7b1f46cd
2026-03-25 08:22:38 +00:00
copilot-swe-agent[bot]