Files

pqhuy1987 e27d877172 [CLAUDE] Infra: adopt AI_INFRA Harness 1/2/3 — monitor subs (roster 8->10) + wave-folder isolation + email channel (S49)

Harness 1 (Self-observability): NEW tooling-auditor (H1 tooling/docs-freshness 4-faceted) + harvest-curator (H2 harvest-integrity 5-axis) INFORM-only monitor subs, TACH BIET per anh-mandate -> roster 8->10. Wire session-start Phase 2.1.1 RE-REPORT + session-end L.b 6->7-step (H2 5-axis GATE + H1 chot + B5 wave-gom). H3 plugin/skill = gop-vai doc, 0 new agent.

Harness 2 (wave-folder isolation): hmw.js WAVE-MODE (subMdPath + tool-aware writeGuard) + .gitignore wave-*/ + agent-teams/ (B6 git-check-ignore verified) + NEW workflows/README convention. Harness 3 (email channel): broadcasts/ (6+6 folder + 13 .gitkeep + _index + inbox/README, committed) + send/check-email cmd (self=se) + adap-apply base-path fix outbox/all/.

HMW-mode ON: recon fan-out 4 read-only agent -> em main single-writer WRITE -> reviewer PASS all 3. Containment: git-diff 1 benign self-MEMORY + chunk-count 2414=2414 (0 RAG-write). Nac executed-file, verified-runtime PENDING CLI restart. 3 adap-reports + session log. Test 181 unchanged (no product code). CI runs (hmw.js/.gitignore/.gitkeep not path-ignored) but no bundle/migration change.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-07 23:18:43 +07:00

4.3 KiB

Raw Blame History

Investigator-API Agent — Persistent Memory

Persistent diary cross-session. Auto-injected first ~200 lines at spawn (L1 HOT). Update BEFORE every stop. Tiered Memory v1: L1 HOT soft-cap ~30KB · L2 archive/ on-demand · L3 RAG search_memory just-in-time. Keep entry ≤ 1.5K chars (gotcha #53). NEW agent S39 (2026-05-29) — split từ investigator (external research half). Internal audit history ở investigator-codebase.

🎯 Role baseline

Read-only EXTERNAL research specialist. WebFetch/WebSearch official docs + NuGet/npm CVE + FE lib eval + cross-project reference (NamGroup/DH_Y_DUOC/BVAAU). Tools: Read, Bash, WebFetch, WebSearch + 5 RAG MCP. Output: findings + source URLs ≤ 500 words. Skill: dependency-audit-erp.

🚫 Split boundary

✅ MINE: external docs, CVE scan, lib license+bundle eval, cross-project pattern port
❌ NOT: internal codebase audit / SQL schema / grep symbol → investigator-codebase

🔑 Dependency pin constraints (flag violation)

MediatR 12.4.1 (gotcha #1) · Swashbuckle 6.9.0 (gotcha #2) · Node CI 20.x · LibreOffice 25.8.6 · @microsoft/signalr 8.0.7
New dep eval: license MIT? bundle gzipped impact? (vd FullCalendar v6 React MIT verified S36 — daygrid/timegrid OK, Premium chỉ scheduler — SOL chọn custom HTML grid save ~80KB instead)

🌐 Trusted source URLs

anthropic.com/engineering/ · cognition.ai/blog/ · learn.microsoft.com/en-us/aspnet/core/ + /ef/core/ · tanstack.com/query/latest · ui.shadcn.com · philschmid.de · eugeneyan.com · hamel.dev

📂 Cross-project reference paths

NamGroup D:\Dropbox\CONG_VIEC\NAMGROUP\SOURCECODE_CÔNG_TY\NAMGROUP\ (Phase 10 port — TblNhanVien/TblPhongHop/TblResource patterns)
DH_Y_DUOC D:\Dropbox\CONG_VIEC\DAI_Y_DUOC\DH_Y_DUOC_SOURCECODE\DH_Y_DUOC\ (clean arch + CQRS)
BVAAU D:\Dropbox\CONG_VIEC\BENHVIEN_A_AU\SOURCE_CODDE\ (multi-agent config — 7 agent split reference)

📅 Recent activity (last 10 FIFO)

2026-05-29 (S39 agent split setup): NEW agent created từ split investigator. Seeded external-research half. Prior cross-project audits (NamGroup Phase 10 port G-H1/G-O2 + FullCalendar eval S36 + BVAAU 7-agent config S39) absorbed into role baseline.
2026-05-29 (S40 FIRST SPAWN — smoke-verify + RAG fleet report): Agent load OK confirmed. list_projects → 7 project, total 39,798 chunks. Rerank pipeline LIVE verdict PASS (search_memory scope=self use_rerank=true → top rerank_score 0.8789, 3 results all carry rerank_score). Staleness >5d (vs 05-29): dh_y_duoc (05-23, 6d) / namgroup_main (05-22, 7d) / ashico_erp (05-22, 7d). solution_erp 05-28 fresh-ish but missing S37-S39 content. shared_global = 0 chunks (chưa promote pattern nào). MINOR drift: namgroup_main actual 11306 (brief said 11305). vipix_ai_infra (1652) = AI_INFRA hub root D:\...\AI_INFRA. No re-ingest performed (report-only).
2026-06-07 (Harness 1 H3 — plugin/skill adoption audit): Browsed marketplace claude-plugins-official (full marketplace.json ~200+ plugin; local folder 35). Enabled user-global (~/.claude/settings.json) = 15 plugin. KEY surprise: sql-database-assistant + frontend-design + skill-creator + code-reviewer exist as standalone user-global skills at ~/.claude/skills/ (auto-available every project — NO plugin enable needed). ~/.claude/skills/ has 23 standalone skills. Value-locus verdict: frontend-design=skill-only (clean); code-modernization=agent-bearing (5 agent incl security-auditor/test-engineer + 7 cmd — NOT skill-only); pr-review-toolkit + feature-dev = agent-bearing, BOTH ship agent code-reviewer → name-collision ×2 + collides roster reviewer.md. csharp-lsp = Windows no-op (csharp-ls NOT on PATH, needs dotnet tool install). code-review = command-only, gh-CLI based → partial no-op (project = Gitea not GitHub). session-report = node .mjs (Win-OK, Node 20). Roster ACTUAL = 8 agent (cicd-monitor/frontend-designer/implementer-backend|frontend/investigator-api|codebase/reviewer/test-specialist); "8→10" = planned H1 tooling-auditor + H2 harvest-curator. Recommend: GỘP skill vào sub hiện-có, KHÔNG enable agent-bearing plugin (roster đủ). Report-only.

🔄 Curate trigger

Size > ~30KB → archive to L2 archive/<period>.md. Stale > 3 months → remove.

4.3 KiB Raw Blame History Unescape Escape