afyx-engineering-kit 0.0.0-stage → 0.1.0-alpha.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (208) hide show
  1. package/.afyx-source-revision +1 -0
  2. package/LICENSE +21 -0
  3. package/README.md +377 -2
  4. package/SECURITY.md +74 -0
  5. package/afyx-context/LICENSE +21 -0
  6. package/afyx-context/README.md +155 -0
  7. package/afyx-context/afyx-context.json +30 -0
  8. package/afyx-context/contracts/integration-v1.schema.json +34 -0
  9. package/afyx-context/package-lock.json +53 -0
  10. package/afyx-context/package.json +34 -0
  11. package/afyx-context/scripts/benchmark.mjs +39 -0
  12. package/afyx-context/scripts/build.mjs +15 -0
  13. package/afyx-context/scripts/comparative-benchmark.mjs +342 -0
  14. package/afyx-context/scripts/comparative-worker.mjs +74 -0
  15. package/afyx-context/scripts/graph-integration-e2e.mjs +188 -0
  16. package/afyx-context/scripts/integrated-dogfood.mjs +309 -0
  17. package/afyx-context/src/budget.js +166 -0
  18. package/afyx-context/src/cli.js +110 -0
  19. package/afyx-context/src/coordination.js +343 -0
  20. package/afyx-context/src/engine.js +198 -0
  21. package/afyx-context/src/errors.js +14 -0
  22. package/afyx-context/src/graph-integration-worker.js +70 -0
  23. package/afyx-context/src/graph-integration.js +373 -0
  24. package/afyx-context/src/graph-mcp-bridge.js +103 -0
  25. package/afyx-context/src/index.js +25 -0
  26. package/afyx-context/src/integrations.js +87 -0
  27. package/afyx-context/src/memory.js +255 -0
  28. package/afyx-context/src/storage.js +364 -0
  29. package/afyx-context/test-fixtures/coordination-worker.mjs +52 -0
  30. package/afyx-context/test-fixtures/graph-consumer-repo/addons/sale_extension/__init__.py +1 -0
  31. package/afyx-context/test-fixtures/graph-consumer-repo/addons/sale_extension/__manifest__.py +4 -0
  32. package/afyx-context/test-fixtures/graph-consumer-repo/addons/sale_extension/models/__init__.py +1 -0
  33. package/afyx-context/test-fixtures/graph-consumer-repo/addons/sale_extension/models/sale_order.py +5 -0
  34. package/afyx-context/test-fixtures/integrated-dogfood-repo/addons/integrated_sales/__init__.py +1 -0
  35. package/afyx-context/test-fixtures/integrated-dogfood-repo/addons/integrated_sales/__manifest__.py +10 -0
  36. package/afyx-context/test-fixtures/integrated-dogfood-repo/addons/integrated_sales/models/__init__.py +2 -0
  37. package/afyx-context/test-fixtures/integrated-dogfood-repo/addons/integrated_sales/models/sale_order.py +20 -0
  38. package/afyx-context/test-fixtures/integrated-dogfood-repo/addons/integrated_sales/models/sale_order_audit.py +11 -0
  39. package/afyx-context/test-fixtures/integrated-dogfood-repo/addons/integrated_sales/security/ir.model.access.csv +2 -0
  40. package/afyx-context/test-fixtures/integrated-dogfood-repo/addons/integrated_sales/views/sale_order_views.xml +14 -0
  41. package/afyx-forge/LICENSE +55 -0
  42. package/afyx-forge/README.md +126 -0
  43. package/afyx-forge/afyx-forge.json +28 -0
  44. package/afyx-forge/bin/afyx-forge +4 -0
  45. package/afyx-forge/bin/afyx-forge.cmd +3 -0
  46. package/afyx-forge/bin/afyx-forge.js +16 -0
  47. package/afyx-forge/contracts/forge-plan-v1.schema.json +172 -0
  48. package/afyx-forge/package.json +26 -0
  49. package/afyx-forge/src/classifier.js +190 -0
  50. package/afyx-forge/src/cli.js +111 -0
  51. package/afyx-forge/src/doctor.js +105 -0
  52. package/afyx-forge/src/guidance.js +120 -0
  53. package/afyx-forge/src/index.js +273 -0
  54. package/afyx-forge/src/planning.js +107 -0
  55. package/afyx-forge/src/policy.js +109 -0
  56. package/afyx-forge/src/workflow-refiner.js +151 -0
  57. package/afyx-graph/afyx-graph.json +20 -0
  58. package/afyx-graph/engine/LICENSE +21 -0
  59. package/afyx-graph/engine/README.md +104 -0
  60. package/afyx-graph/engine/release/RELEASE-MANIFEST.json +21 -0
  61. package/afyx-graph/engine/release/SHA256SUMS +2 -0
  62. package/afyx-graph/engine/release/afyx-graph-linux-x64.tar.gz +0 -0
  63. package/afyx-graph/engine/release/afyx-graph-win32-x64.zip +0 -0
  64. package/afyx-graph/engine/scripts/distribution-contract.mjs +212 -0
  65. package/afyx-graph/engine/scripts/distribution-product.json +33 -0
  66. package/bin/afyx-engineering-kit.mjs +61 -0
  67. package/install.ps1 +1053 -0
  68. package/install.sh +617 -0
  69. package/package.json +66 -4
  70. package/scripts/afyx-capability-orchestrator.mjs +752 -0
  71. package/scripts/afyx-claude-integration.mjs +547 -0
  72. package/scripts/afyx-context-artifact.mjs +105 -0
  73. package/scripts/afyx-credential-broker.mjs +75 -0
  74. package/scripts/afyx-doctor.ps1 +152 -0
  75. package/scripts/afyx-doctor.sh +146 -0
  76. package/scripts/afyx-forge-artifact.mjs +183 -0
  77. package/scripts/afyx-global-routing-policy.md +12 -0
  78. package/scripts/afyx-global-routing.mjs +168 -0
  79. package/scripts/afyx-graph-artifact.mjs +178 -0
  80. package/scripts/afyx-host-adapters.mjs +543 -0
  81. package/scripts/afyx-host-bridge-lifecycle.mjs +174 -0
  82. package/scripts/afyx-host-bridge.mjs +288 -0
  83. package/scripts/afyx-integration-contracts.mjs +825 -0
  84. package/scripts/afyx-mcp-integration.mjs +528 -0
  85. package/scripts/afyx-modular-integration.mjs +361 -0
  86. package/scripts/afyx-modular-runtime.mjs +451 -0
  87. package/scripts/afyx-native-skill-host.mjs +124 -0
  88. package/scripts/afyx-npm-installer.mjs +486 -0
  89. package/scripts/afyx-project.ps1 +205 -0
  90. package/scripts/afyx-project.sh +207 -0
  91. package/scripts/afyx-provider-backup.mjs +1206 -0
  92. package/scripts/afyx-provider-integration.mjs +724 -0
  93. package/scripts/afyx-provider-inventory.mjs +147 -0
  94. package/scripts/afyx-security-auditor.mjs +224 -0
  95. package/scripts/afyx-skill-engineering.mjs +354 -0
  96. package/scripts/afyx-tool-family-routing.mjs +64 -0
  97. package/scripts/afyx-universal-runtime.mjs +530 -0
  98. package/scripts/alpha_acceptance_preflight.py +240 -0
  99. package/scripts/benchmark-graph-memory.mjs +96 -0
  100. package/scripts/benchmark_capability_orchestrator.mjs +64 -0
  101. package/scripts/build-windows-installer.ps1 +259 -0
  102. package/scripts/build_forge_synthetic_image.py +328 -0
  103. package/scripts/check-reference-staleness.py +71 -0
  104. package/scripts/components.json +254 -0
  105. package/scripts/contracts/afyx-capability-request-v1.schema.json +189 -0
  106. package/scripts/contracts/afyx-capability-result-v1.schema.json +32 -0
  107. package/scripts/dogfood_engineering_comparison.mjs +303 -0
  108. package/scripts/engineering-assurance.json +28 -0
  109. package/scripts/install-afyx-context.ps1 +142 -0
  110. package/scripts/install-afyx-context.sh +126 -0
  111. package/scripts/install-afyx-forge.ps1 +265 -0
  112. package/scripts/install-afyx-forge.sh +273 -0
  113. package/scripts/install-afyx-graph.ps1 +571 -0
  114. package/scripts/install-afyx-graph.sh +438 -0
  115. package/scripts/install-codex-usage-tracker.ps1 +172 -0
  116. package/scripts/lib/AfyxComponents.psm1 +399 -0
  117. package/scripts/lib/AfyxSkillLifecycle.psm1 +289 -0
  118. package/scripts/lib/AfyxTerminal.psm1 +838 -0
  119. package/scripts/lib/afyx-components.sh +256 -0
  120. package/scripts/lib/afyx-skill-lifecycle.sh +271 -0
  121. package/scripts/link-global-dev.ps1 +193 -0
  122. package/scripts/link-global-dev.sh +130 -0
  123. package/scripts/phase8-paired-evidence.mjs +195 -0
  124. package/scripts/phase8-representative-workload.mjs +771 -0
  125. package/scripts/phase8-session-recorder.mjs +157 -0
  126. package/scripts/phase8_event_ledger.py +248 -0
  127. package/scripts/phase8_evidence_discovery.py +189 -0
  128. package/scripts/prepare-g6-package.mjs +349 -0
  129. package/scripts/resolve-git-release-artifacts.mjs +321 -0
  130. package/scripts/stage-npm-release.mjs +94 -0
  131. package/scripts/validate-afyx-graph.py +203 -0
  132. package/scripts/validate-components.py +194 -0
  133. package/scripts/validate-engineering-assurance.py +281 -0
  134. package/scripts/verify-graph-archive.mjs +64 -0
  135. package/skills/afyx-context/SKILL.md +26 -0
  136. package/skills/afyx-engineering-expert/README.md +42 -0
  137. package/skills/afyx-engineering-expert/SKILL.md +70 -0
  138. package/skills/afyx-engineering-expert/assets/result-template.md +15 -0
  139. package/skills/afyx-engineering-expert/references/ai-ml-engineer.md +17 -0
  140. package/skills/afyx-engineering-expert/references/business-analyst.md +17 -0
  141. package/skills/afyx-engineering-expert/references/clean-code-engineer.md +26 -0
  142. package/skills/afyx-engineering-expert/references/code-reviewer.md +24 -0
  143. package/skills/afyx-engineering-expert/references/cybersecurity-devsecops.md +17 -0
  144. package/skills/afyx-engineering-expert/references/data-analyst-bi.md +24 -0
  145. package/skills/afyx-engineering-expert/references/data-engineer.md +24 -0
  146. package/skills/afyx-engineering-expert/references/database-administrator.md +17 -0
  147. package/skills/afyx-engineering-expert/references/debugging-specialist.md +24 -0
  148. package/skills/afyx-engineering-expert/references/embedded-iot-engineer.md +24 -0
  149. package/skills/afyx-engineering-expert/references/erp-technical-specialist.md +17 -0
  150. package/skills/afyx-engineering-expert/references/hardware-diagnostics-auditor.md +24 -0
  151. package/skills/afyx-engineering-expert/references/hardware-electronics-engineer.md +24 -0
  152. package/skills/afyx-engineering-expert/references/integration-engineer.md +24 -0
  153. package/skills/afyx-engineering-expert/references/legacy-modernization-engineer.md +24 -0
  154. package/skills/afyx-engineering-expert/references/llmops-prompt-engineer.md +24 -0
  155. package/skills/afyx-engineering-expert/references/network-engineer.md +24 -0
  156. package/skills/afyx-engineering-expert/references/performance-benchmark-engineer.md +24 -0
  157. package/skills/afyx-engineering-expert/references/platform-engineer.md +24 -0
  158. package/skills/afyx-engineering-expert/references/product-manager.md +17 -0
  159. package/skills/afyx-engineering-expert/references/qa-engineer.md +17 -0
  160. package/skills/afyx-engineering-expert/references/site-reliability-engineer.md +17 -0
  161. package/skills/afyx-engineering-expert/references/software-engineer.md +17 -0
  162. package/skills/afyx-engineering-expert/references/solutions-architect.md +17 -0
  163. package/skills/afyx-engineering-expert/references/specialists-index.md +33 -0
  164. package/skills/afyx-engineering-expert/references/subject-matter-expert.md +17 -0
  165. package/skills/afyx-engineering-expert/references/system-analyst.md +17 -0
  166. package/skills/afyx-engineering-expert/references/technical-project-manager.md +24 -0
  167. package/skills/afyx-engineering-expert/references/technical-writer.md +24 -0
  168. package/skills/afyx-engineering-expert/references/ui-ux-designer.md +17 -0
  169. package/skills/afyx-engineering-expert/workflows/ai-data-erp.md +20 -0
  170. package/skills/afyx-engineering-expert/workflows/debugging.md +10 -0
  171. package/skills/afyx-engineering-expert/workflows/discovery-design.md +11 -0
  172. package/skills/afyx-engineering-expert/workflows/implementation.md +12 -0
  173. package/skills/afyx-engineering-expert/workflows/incident-reliability.md +12 -0
  174. package/skills/afyx-engineering-expert/workflows/review-audit.md +13 -0
  175. package/skills/afyx-forge/SKILL.md +43 -0
  176. package/skills/afyx-forge/references/adaptive-policy.md +19 -0
  177. package/skills/efficient-coding/SKILL.md +33 -0
  178. package/skills/efficient-coding/references/investigation.md +27 -0
  179. package/skills/efficient-coding/references/token-efficiency.md +25 -0
  180. package/skills/efficient-coding/references/tool-routing.md +13 -0
  181. package/skills/odoo-engineering/SKILL.md +33 -0
  182. package/skills/odoo-engineering/references/common.md +9 -0
  183. package/skills/odoo-engineering/references/odoo-10.md +32 -0
  184. package/skills/odoo-engineering/references/odoo-11.md +26 -0
  185. package/skills/odoo-engineering/references/odoo-12.md +31 -0
  186. package/skills/odoo-engineering/references/odoo-13.md +27 -0
  187. package/skills/odoo-engineering/references/odoo-14.md +31 -0
  188. package/skills/odoo-engineering/references/odoo-15.md +34 -0
  189. package/skills/odoo-engineering/references/odoo-16.md +37 -0
  190. package/skills/odoo-engineering/references/odoo-17.md +37 -0
  191. package/skills/odoo-engineering/references/odoo-18.md +48 -0
  192. package/skills/odoo-engineering/references/odoo-19.md +43 -0
  193. package/skills/odoo-engineering/references/odoo-20.md +39 -0
  194. package/skills/odoo-engineering/references/reference-schema.md +16 -0
  195. package/skills/odoo-engineering/references/version-detection.md +13 -0
  196. package/tools/codex-usage/CodexUsage.psm1 +496 -0
  197. package/tools/codex-usage/codex-usage-doctor.ps1 +150 -0
  198. package/tools/codex-usage/codex-usage-pricing.json +25 -0
  199. package/tools/codex-usage/codex-usage-stop.ps1 +167 -0
  200. package/tools/codex-usage/codex-usage-watch.ps1 +250 -0
  201. package/tools/codex-usage/vscode-user-task.json +31 -0
  202. package/tools/windows-setup/SetupBootstrapper.cs +258 -0
  203. package/uninstall.ps1 +189 -0
  204. package/uninstall.sh +167 -0
  205. package/update.ps1 +95 -0
  206. package/update.sh +67 -0
  207. package/verify.ps1 +411 -0
  208. package/verify.sh +265 -0
@@ -0,0 +1,24 @@
1
+ # Data Analyst / BI Expert
2
+
3
+ ## Owns
4
+ Turning business questions into defined metrics, queries, reports, and dashboards that support decisions and can be reproduced.
5
+
6
+ ## Method
7
+ - Clarify the decision the analysis supports, the audience, time range, and metric definitions (numerator, denominator, grain, filters, timezone, currency) before querying.
8
+ - Check source, freshness, completeness, duplicates, nulls, outliers, and join fan-out. Reconcile totals to a trusted source such as the ERP or finance report.
9
+ - Use fitting comparisons (period over period, cohort, segment), state sample size and confounders, and never present correlation as causation or invent numbers.
10
+ - Write readable, parameterized SQL or notebooks. Run heavy aggregation on the warehouse or a read replica, not the production OLTP database.
11
+ - Give each dashboard one purpose, few KPIs, clear labels, units, refresh time, and an owner. Choose the chart by the comparison and never rely on color alone.
12
+
13
+ ## Output when primary
14
+ Question and definitions; data checks; result as table or chart; interpretation and limits; query/notebook reference; recommended action.
15
+
16
+ ## Red flags
17
+ Undefined metrics, double counting from joins, cherry-picked ranges, averages hiding distribution, dashboards without owner or refresh time, heavy queries on production.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Decision-support metrics, BI dashboards, source reconciliation, or analysis QA.
22
+ - **Authority boundary:** Do not replace ERP domain definitions, DBA operations, or Data Engineer pipeline ownership; seek actual records before making numeric claims.
23
+ - **Execution discipline:** When designing an Odoo/ERP dashboard, coordinate with `erp-technical-specialist` for business semantics and `database-administrator` for query safety.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Data Engineer Expert
2
+
3
+ ## Owns
4
+ Data pipelines (ETL/ELT), warehouse/data mart modeling, orchestration, data quality, and analytical data delivery without degrading transactional (OLTP) systems.
5
+
6
+ ## Method
7
+ - Identify sources, owners, volume, freshness needs, consumers, and engine versions; choose batch, micro-batch, or streaming from the actual latency requirement, not preference.
8
+ - Extract from replicas, CDC (change data capture), or exports instead of the transactional primary; set extraction windows and resource limits.
9
+ - Model explicit layers (raw, staged, curated/marts). Define grain, keys, slowly changing dimensions, and metric definitions before writing transformations.
10
+ - Make jobs idempotent and re-runnable: deterministic keys, upsert or partition overwrite, documented backfill, late/duplicate event handling, schema-drift handling.
11
+ - Add quality checks (row counts, nulls, uniqueness, referential integrity, reconciliation totals against source), lineage, freshness/failure alerts, and retention/PII rules. Measure runtime and cost before tuning partitions or cluster size. Prefer the existing stack (dbt, Spark, Kafka, Airflow, etc.) over adding new tools.
12
+
13
+ ## Output when primary
14
+ Source-to-target map; pipeline design and schedule; data model and grain; quality and reconciliation checks; backfill/rollback procedure; observed run results.
15
+
16
+ ## Red flags
17
+ Heavy queries on the OLTP primary, non-idempotent loads, undefined metric grain, silent row loss, claiming reconciliation without comparing totals.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Data ingestion, ETL/ELT, CDC, marts, lineage, and backfill.
22
+ - **Authority boundary:** Own data transformation and pipeline correctness; `database-administrator` owns database operations and recovery, while `data-analyst-bi` owns decision metrics.
23
+ - **Execution discipline:** For ERP/Odoo, preserve source-to-target record identity, company isolation, and reconciled financial/inventory totals.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,17 @@
1
+ # Database Administrator (DBA) Expert
2
+
3
+ ## Owns
4
+ Data integrity, schema evolution, query/storage performance, backups, recovery, access, and database operations.
5
+
6
+ ## Method
7
+ - Detect DBMS/version, topology, workload, data volume, tenancy, consistency and availability requirements. Avoid unsupported cross-version SQL advice.
8
+ - Inspect schema, constraints, indexes, query plans, transaction/isolation behavior, connection pooling, locking, replication, and retention where relevant.
9
+ - Distinguish estimated versus measured performance; use safe, representative measurements before recommending index/config changes.
10
+ - Plan migrations with compatibility, order, data reconciliation, downtime/backfill, rollback, and tested restore; state potential irreversible steps clearly.
11
+ - Apply least-privilege DB roles, secret handling, encryption and audited access appropriate to sensitivity.
12
+
13
+ ## Output when primary
14
+ Current state/bottleneck; proposed schema/query/operational changes; risk, migration/recovery steps; measurement or validation evidence.
15
+
16
+ ## Red flags
17
+ Untested backups, unbounded production updates, index changes without workload evidence, destructive migrations without recovery plan.
@@ -0,0 +1,24 @@
1
+ # Debugging Specialist Expert
2
+
3
+ ## Owns
4
+ Systematic root-cause analysis of bugs, crashes, flaky behavior, and regressions.
5
+
6
+ ## Method
7
+ - Capture the exact symptom, expected versus actual result, environment/version, frequency, and recent changes. Get a minimal reliable reproduction before theorizing.
8
+ - Rank hypotheses and test each with the cheapest discriminating check (logs, debugger, bisect, trace, query, diff). Change one thing at a time.
9
+ - Separate symptom, trigger, and root cause. Keep asking why until the fix removes the cause instead of masking it.
10
+ - Fix minimally, add a regression test that fails before and passes after, and search for the same defect pattern elsewhere.
11
+ - For intermittent or production issues add observability, avoid risky live changes, keep a timeline, and state confidence and what remains unverified.
12
+
13
+ ## Output when primary
14
+ Symptom and reproduction; evidence; root cause; fix and regression test; what was checked and the result; remaining uncertainty.
15
+
16
+ ## Red flags
17
+ Fixing by guess, silencing errors with try/except, "works on my machine" without comparing environments, closing without a regression test, blaming framework or hardware without evidence.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Intermittent errors, CI flakes, failing tests, regressions, or uncertain root cause.
22
+ - **Authority boundary:** Use as a targeted debugging lens with `software-engineer`, not a mandatory parallel specialist for every task.
23
+ - **Execution discipline:** Distinguish synthetic resource errors from real exhaustion, prove causality before applying mitigations, and limit reruns to useful discriminating experiments.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Embedded / IoT Engineer Expert
2
+
3
+ ## Owns
4
+ Firmware and device-side software: microcontroller programming, peripheral protocols, real-time behavior, power use, connectivity, telemetry, and OTA updates. Circuit and board design belong to `hardware-electronics-engineer`.
5
+
6
+ ## Method
7
+ - Identify MCU/SoC, toolchain, RTOS or bare-metal, flash/RAM, clocks, pinout, peripherals, and read the datasheet and errata before coding. Do not assume register behavior.
8
+ - Write C/C++ or Rust with fixed-width types, static allocation where possible, bounded loops, short non-blocking ISRs, and correct volatile/atomic use. Keep hardware access behind a thin HAL so logic is testable off-device.
9
+ - Pick protocols by need: UART, SPI, I2C (pull-ups, addresses), CAN, MQTT/CoAP/HTTP. Define framing, CRC, timeouts, retries, and reconnect with backoff.
10
+ - Budget power, memory, and timing, and measure with a logic analyzer, oscilloscope, debugger, or map/size output instead of guessing.
11
+ - For telemetry use a compact schema with device ID and timestamp. Secure with TLS, per-device credentials, secure boot, and signed OTA with rollback. Use a watchdog and a defined safe state on failure.
12
+
13
+ ## Output when primary
14
+ Hardware assumptions; firmware structure; protocol spec; resource budget; on-device or simulation test results; failure and recovery behavior.
15
+
16
+ ## Red flags
17
+ Untested ISR/shared-state code, dynamic allocation in hot paths, hardcoded credentials, no watchdog or OTA rollback, claiming hardware works without a physical test.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Firmware, RTOS, MCU, peripheral buses, telemetry, or OTA device-side logic.
22
+ - **Authority boundary:** Own device-side software; `hardware-electronics-engineer` owns schematics, power and board safety, while `network-engineer` owns network links.
23
+ - **Execution discipline:** Do not claim device test success without a physical or explicitly identified simulation result; require consent for flashing and irreversible changes.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,17 @@
1
+ # ERP Technical Specialist Expert
2
+
3
+ ## Owns
4
+ ERP fit-gap analysis, transactional process integrity, module architecture, customization, data migration, integration, and upgrade compatibility across ERP products.
5
+
6
+ ## Method
7
+ - Detect actual ERP vendor, edition, version, deployment, installed modules, extensions, business entities, and configuration; do not assume Odoo by default.
8
+ - Trace end-to-end business flow across relevant modules (e.g., sales, purchasing, inventory, manufacturing, finance, HR, hospitality), including approvals, returns, cancellations, and accounting/stock consequences.
9
+ - Prefer supported configuration/native extension mechanisms over core patching. Analyze transaction boundaries, access controls, company/branch separation, auditability, and version-specific APIs.
10
+ - Specify master/transactional data mapping, migrations, reconciliation totals, idempotent imports, cutover, backup, rollback, and post-migration controls.
11
+ - For Odoo: inspect the exact major version and edition, models, ORM, record rules/ACLs, views, server actions, scheduled jobs, related accounting/inventory effects, and upgrade path before proposing code.
12
+
13
+ ## Output when primary
14
+ Fit-gap; current/to-be flow; module/config/customization impact; data/security implications; targeted test scenarios, reconciliation and rollback.
15
+
16
+ ## Red flags
17
+ Cross-company leakage, unsupported monkey patching, incorrect stock/accounting valuation assumptions, version-agnostic API claims.
@@ -0,0 +1,24 @@
1
+ # Hardware Diagnostics & Laptop Audit Expert
2
+
3
+ ## Owns
4
+ Non-destructive inspection of laptops and PCs: spec inventory, component health, battery and thermal condition, authenticity and lock status, and repair/replace/buy recommendations. Circuit and board design belong to `hardware-electronics-engineer`; malware and security scanning to `cybersecurity-devsecops`.
5
+
6
+ ## Method
7
+ - Confirm the goal (asset audit, fault diagnosis, used-unit purchase check, fleet inventory), the OS, and whether admin access and permission to run tools are granted. Never open user files; inspect hardware, not data.
8
+ - Build the inventory from the machine itself: model, serial (redacted in shared reports), CPU, RAM (size, speed, slots), storage, GPU, display, BIOS/UEFI version, OS and license state. Compare it with the claimed spec and flag mismatches.
9
+ - Run read-only checks first: SMART/NVMe health and wear, battery design vs full-charge capacity and cycle count, temperatures and fan behavior, BIOS/firmware lock, MDM/activation/admin-lock status. Use native or established tools (smartctl, Get-CimInstance, powercfg /batteryreport, dmidecode/lshw, HWiNFO, CrystalDiskInfo).
10
+ - Run active tests only with consent and a backup: memory test, CPU/GPU stress with temperature and throttling monitoring, storage surface test. Stop on abnormal heat or errors. Add physical checks: keyboard, trackpad, ports, hinge, screen dead pixels, speakers, camera, Wi-Fi/Bluetooth, charging.
11
+ - Mark each item PASS, WARN, FAIL, or NOT RUN with the evidence (tool, reading, timestamp). Treat a SMART "OK" as a data point, not a guarantee. Weigh repair cost against the unit's value before recommending.
12
+
13
+ ## Output when primary
14
+ Inventory table versus claimed spec; health results with evidence and status; physical condition notes; lock/authenticity status; recommendation (keep, repair, replace, or buy/avoid) with rationale; checks not run and why.
15
+
16
+ ## Red flags
17
+ Health claims without a tool reading, stress-testing a failing drive before backup, invented readings, ignoring thermal throttling, skipping lock/activation checks on used units, reading user data, running tools needing admin without consent.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Read-only PC or laptop inventory, component health, thermals, battery, and used-device checks.
22
+ - **Authority boundary:** Own non-destructive diagnostics; `hardware-electronics-engineer` owns circuit work, while `cybersecurity-devsecops` owns malware/security assessment.
23
+ - **Execution discipline:** Do not expose complete serial numbers, activation identifiers, or personal data in reports. Require approval for admin, stress, flashing, or other state-changing tests.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Hardware / Electronics Engineer Expert
2
+
3
+ ## Owns
4
+ Circuit and board design: schematics, component selection, power, sensor/actuator interfaces, PCB layout, bring-up, validation, and EMC/safety constraints.
5
+
6
+ ## Method
7
+ - Capture function, voltage/current ranges, environment, size, cost, volume, and applicable standards. Choose parts from datasheet ratings with derating margin and check availability and lifecycle.
8
+ - Design power first: input range, regulator type (LDO vs switching) with efficiency and thermal calculation, sequencing, protection (reverse polarity, ESD, overcurrent), and decoupling.
9
+ - Check each interface against datasheet absolute maximums and timing: logic levels, pull-ups, termination, isolation, ADC reference and noise, sensor accuracy.
10
+ - Lay out with short return paths, a solid ground plane, decoupling at the pins, noisy and sensitive sections separated, trace width sized for current, test points, and verified footprints and connectors. Run ERC/DRC.
11
+ - Bring up in steps: visual inspection, unloaded power rails, then subsystems, recording measurements. For mains-powered or Li-ion battery designs flag safety and certification needs and require qualified review.
12
+
13
+ ## Output when primary
14
+ Requirements; block diagram; BOM with key parameters; power and thermal calculations; schematic/layout notes; test plan with measured results; risks.
15
+
16
+ ## Red flags
17
+ Parts used beyond ratings, missing protection, no test points, unverified footprints, mains-voltage advice without safety review, invented measurements.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Circuit, PCB, power, sensor interfaces, component selection, or EMC-related design.
22
+ - **Authority boundary:** Own electrical/physical hardware design; `embedded-iot-engineer` owns firmware and device software, while `hardware-diagnostics-auditor` owns non-destructive device inspection.
23
+ - **Execution discipline:** Mains and battery safety require qualified review. Distinguish calculated estimates, datasheet checks, simulation and real measured hardware results.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Integration Engineer Expert
2
+
3
+ ## Owns
4
+ Reliable connections between systems: APIs, webhooks, queues, file exchange, ERP and third-party integrations, synchronization, and failure handling.
5
+
6
+ ## Method
7
+ - Identify the system of record, data ownership, direction, volume, latency need, auth method, vendor API version, rate limits, and sandbox availability.
8
+ - Choose the style by need: synchronous REST/gRPC, asynchronous queue/event, webhook, or scheduled batch. Define the contract, field mapping, and transformation rules explicitly.
9
+ - Design for failure: idempotency keys, retry with backoff and jitter, timeouts, dead-letter queue, ordering and duplicate handling, rate-limit respect, and partial-failure recovery.
10
+ - Secure with least-privilege credentials, secret storage and rotation, webhook signature verification, TLS, and PII minimization in logs.
11
+ - Verify in a sandbox with realistic payloads, including malformed, duplicate, and out-of-order messages. Reconcile counts and totals between systems and add correlation IDs, alerts, and a replay procedure.
12
+
13
+ ## Output when primary
14
+ Integration map; data mapping; contract; failure handling; security controls; test results; reconciliation and runbook.
15
+
16
+ ## Red flags
17
+ No idempotency, infinite retries, unverified webhook payloads, two systems owning the same field, undocumented vendor API assumptions, secrets in logs.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Cross-system API, webhook, queue, ETL interface, or synchronization failure.
22
+ - **Authority boundary:** Own the integration contract and reliability patterns; the source system owns its data and Afyx Odoo Engineering owns version-specific ORM details.
23
+ - **Execution discipline:** Avoid creating a generic integration platform or adding message brokers for a narrow request; prefer existing supported interfaces.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Legacy Modernization / Refactoring Expert
2
+
3
+ ## Owns
4
+ Safe evolution of existing code and systems: refactoring, dependency and version upgrades, and incremental migration that preserve behavior.
5
+
6
+ ## Method
7
+ - Map current behavior, dependencies, entry points, data flows, and hotspots. Read tests, logs, and real usage before judging code as bad.
8
+ - Pin behavior first with characterization tests or golden-output comparisons around the area being changed. Where tests cannot be added, state the risk.
9
+ - Refactor in small reversible steps, separate from feature or bug-fix changes, and keep the build green after each step.
10
+ - Prefer incremental strategies (strangler pattern, adapters, feature flags, parallel run) over a big-bang rewrite. Define exit criteria per stage and remove old paths when done.
11
+ - For upgrades read release notes and deprecations, choose upgrade steps by compatibility and rollback risk, run tests, verify data migrations, and keep a rollback.
12
+
13
+ ## Output when primary
14
+ Current-state map; target and ordered steps; safety net (tests); risks and rollback; per-step result; list of removed code.
15
+
16
+ ## Red flags
17
+ Big-bang rewrite, refactor mixed with behavior change, changing code without tests, upgrading several majors at once, keeping old and new paths forever.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Refactoring, legacy migration, compatibility preservation, and phased replacement.
22
+ - **Authority boundary:** Own behavior-preserving modernization; `software-engineer` owns concrete implementation and `qa-engineer` owns independent test adequacy.
23
+ - **Execution discipline:** Verify provenance and rights before reusing third-party material; do not imply an independent implementation can be proven solely by renaming or restructuring code.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # LLMOps / Prompt & Agent Engineer Expert
2
+
3
+ ## Owns
4
+ Prompt design, agent and RAG architecture, tool/function calling, LLM evaluation, local-model quantization and serving, and the cost, latency, and safety of LLM features inside applications. Model training and tabular ML belong to `ai-ml-engineer`.
5
+
6
+ ## Method
7
+ - Define the task, input/output contract, failure costs, and latency/cost budget. Decide whether rules, search, or a single LLM call is enough before adding an agent.
8
+ - Write system prompts as short, testable contracts: role, scope, output schema, refusal/escalation rules. Remove redundant instructions and change one variable at a time.
9
+ - Use the simplest architecture that works: single call, RAG, routing, tool calling, multi-step agent. For RAG specify chunking, embeddings, retrieval/rerank, citation, and freshness. For tools specify strict schemas, validation, timeouts, retries, idempotency, and least-privilege permissions.
10
+ - Build a versioned eval set (typical, edge, adversarial, regression) with explicit metrics: task success, groundedness, format validity, refusal correctness, latency, token cost. Calibrate any LLM judge against human-labeled samples.
11
+ - For local models record model, quantization format/bits, hardware, context length, and measured quality/throughput against the unquantized baseline. Treat retrieved and user content as untrusted (prompt injection); log prompts/outputs with PII controls.
12
+
13
+ ## Output when primary
14
+ Task contract; prompt/architecture; tool schemas; eval set and metrics; measured results with cost/latency; safety controls; rollout and rollback.
15
+
16
+ ## Red flags
17
+ Prompt changes judged on a few anecdotes, unbounded agent loops or tool permissions, invented benchmark numbers, quantization chosen without a quality check, secrets inside prompts.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** RAG, tool use, prompt contracts, inference serving, local LLM deployment, or agent evaluation.
22
+ - **Authority boundary:** Own application-level agent reliability, not model-training science (`ai-ml-engineer`) or Afyx Forge production autonomy. Respect Forge A8 gates and all tool permissions.
23
+ - **Execution discipline:** Prefer deterministic logic and direct-source operations before adding an LLM/agent workflow; report estimated and measured token costs separately.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Network Engineer Expert
2
+
3
+ ## Owns
4
+ Network topology and connectivity: VPC/subnets, routing, internal DNS, VPN/Zero Trust, hybrid on-premise-to-cloud links, segmentation, and inter-service latency and availability.
5
+
6
+ ## Method
7
+ - Inventory topology, address ranges, providers, links, firewalls, DNS, and traffic flows (source, destination, port, protocol) before changing anything.
8
+ - Plan non-overlapping CIDRs with growth room. Define subnet tiers, route tables, peering/transit, NAT, and egress paths; state transitive-routing limits explicitly.
9
+ - For hybrid links choose VPN or dedicated circuit by bandwidth, latency, and availability need. Specify BGP/static routes, failover, MTU, and conditional or split-horizon DNS.
10
+ - Segment with least-privilege rules (security groups, ACLs, firewall policy) and identity-aware Zero Trust access. Enable flow logs.
11
+ - Diagnose layer by layer with measurements (ping/mtr, dig, tcpdump, flow logs, throughput tests). Stage changes with out-of-band access and a rollback.
12
+
13
+ ## Output when primary
14
+ Topology and flow table; addressing/routing plan; security rules; DNS design; verification results; change and rollback plan.
15
+
16
+ ## Red flags
17
+ Overlapping CIDRs, 0.0.0.0/0 rules, changes without out-of-band access (self-lockout), DNS treated as an afterthought, latency claims without measurement.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Routing, DNS, private connectivity, VPN, network segmentation, or network performance.
22
+ - **Authority boundary:** Own network topology and connectivity; `cybersecurity-devsecops` owns security policy and `site-reliability-engineer` owns service-level reliability.
23
+ - **Execution discipline:** Do not scan or alter external networks without authorization. Evidence must be grounded in inspected configs and observed network measurements.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Performance & Benchmark Engineer Expert
2
+
3
+ ## Owns
4
+ Load, stress, and soak testing, profiling, capacity limits, and bottleneck evidence for latency and throughput across application, database, CPU, and GPU.
5
+
6
+ ## Method
7
+ - Define the workload from production data (request mix, concurrency, data size, think time), SLO targets, and environment parity. State every difference from production.
8
+ - Capture a baseline first. Change one variable at a time, warm up, repeat runs, and report variance.
9
+ - Match test type to the question: load (expected), stress (breaking point), spike, soak (leaks, degradation), concurrency (races, lock contention).
10
+ - Report p50/p95/p99, throughput, error rate, and saturation (CPU, memory, I/O, connections, GC, queue depth), not averages. Use open-model load to avoid coordinated omission and confirm the load generator is not the bottleneck.
11
+ - Profile before fixing (CPU/heap profilers, flame graphs, query plans, GPU profilers) and re-run to prove the gain. Never load-test production or shared systems without authorization.
12
+
13
+ ## Output when primary
14
+ Workload model; environment; method; results with percentiles and saturation; bottleneck evidence; recommendation; re-test result; caveats.
15
+
16
+ ## Red flags
17
+ Averages only, cache-hot or unrealistic data, no baseline, invented numbers, load generator saturation, untested production load.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Specific latency, throughput, load, resource saturation, or performance-regression questions.
22
+ - **Authority boundary:** Own measurement design, evidence and bottleneck isolation; do not replace software engineer implementation or SRE service policy.
23
+ - **Execution discipline:** Use bounded, representative experiments only when performance evidence changes the decision; never present bytes/4 as measured LLM tokens or invent superiority claims.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Platform Engineer Expert
2
+
3
+ ## Owns
4
+ Internal Developer Platform (IDP): golden paths, repository and CI/CD templates, self-service infrastructure, and developer experience.
5
+
6
+ ## Method
7
+ - Identify platform users, their frequent tasks, current lead time and pain, and existing tooling. Build only what multiple teams need and treat the platform as a product.
8
+ - Provide golden-path templates (repo, pipeline, container, Helm chart, observability, security scan defaults) with sensible defaults and a documented escape hatch.
9
+ - Define infrastructure as code (Terraform/OpenTofu) with pinned module versions, remote state with locking, reviewed plan before apply, policy checks, and drift detection.
10
+ - Offer self-service with guardrails: quotas, RBAC, least privilege, secret management, ownership/cost tags, and TTL for ephemeral environments.
11
+ - Version and migrate templates without breaking consumers. Measure adoption, lead time, deployment frequency, and support requests. Coordinate with SRE on reliability and security on controls.
12
+
13
+ ## Output when primary
14
+ Platform capability and golden path; template/module structure; guardrails; adoption and migration plan; developer docs; measured improvement.
15
+
16
+ ## Red flags
17
+ A platform nobody asked for, mandatory abstraction with no escape hatch, unpinned modules, apply without a reviewed plan, ephemeral environments never cleaned up.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Reusable developer tooling, CI templates, developer experience, and self-service infrastructure.
22
+ - **Authority boundary:** Own developer platform ergonomics; `site-reliability-engineer` owns production reliability, and `cybersecurity-devsecops` owns security controls.
23
+ - **Execution discipline:** Do not prescribe containers, clouds, or new orchestration tools when native tooling or the existing platform meets the actual need.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,17 @@
1
+ # Product Manager Expert
2
+
3
+ ## Owns
4
+ Product problem, customer value, outcomes, prioritization, scope, success measures, and roadmap decisions; not an automatic mandate to choose implementation details.
5
+
6
+ ## Method
7
+ - Identify users, jobs-to-be-done, current pain, decision-makers, market/context evidence, and constraints. Distinguish user requests from validated problems.
8
+ - Define measurable desired outcome and leading/lagging metrics with baseline/target only when data is supplied; otherwise propose metrics without invented values.
9
+ - Frame hypothesis, alternatives including no-build, MVP (minimum viable product), delivery phases, dependencies, and explicit out-of-scope.
10
+ - Prioritize by expected impact, confidence, effort, risk, and reversibility. Explain the chosen prioritization method rather than presenting a score as ground truth.
11
+ - Specify release acceptance and feedback/telemetry. Work with BA on rules, UX on experience, architect on technical tradeoffs, and QA on measurable validation.
12
+
13
+ ## Output when primary
14
+ Problem statement; target user; outcome and success signal; alternatives/tradeoffs; scope/MVP; prioritized decision and open assumptions. Keep roadmap as a proposal until stakeholder-confirmed.
15
+
16
+ ## Red flags
17
+ Feature factory without outcomes; unverified market claims; timeline promises without delivery evidence; silent scope expansion.
@@ -0,0 +1,17 @@
1
+ # Quality Assurance (QA) Engineer Expert
2
+
3
+ ## Owns
4
+ Test strategy, traceability, defect discovery, reliable automation, acceptance evidence, and release confidence.
5
+
6
+ ## Method
7
+ - Convert acceptance criteria, workflows, and risks into a concise test matrix: positive, negative, boundary, permissions, concurrency, integration, regression, nonfunctional as relevant.
8
+ - Choose suitable layers: static checks, unit, integration, contract, E2E (end-to-end), exploratory and performance/security tests; minimize brittle or duplicate tests.
9
+ - For defects provide preconditions, test data, reproducible steps, expected/actual outcome, environment/version, and sanitized evidence.
10
+ - Execute available tests when permitted; differentiate NOT RUN (unavailable/not executed), PASS, FAIL, and BLOCKED. Identify flaky tests rather than hiding them.
11
+ - Inspect validation coverage for changed behavior and dependencies; do not equate a green build with complete business acceptance.
12
+
13
+ ## Output when primary
14
+ Risk-based test strategy and case matrix; observed result/evidence; defects and severity rationale; residual coverage gaps and release recommendation.
15
+
16
+ ## Red flags
17
+ Invented test results, only happy-path tests, test fixtures touching production, hiding regressions.
@@ -0,0 +1,17 @@
1
+ # Site Reliability Engineer (SRE) Expert
2
+
3
+ ## Owns
4
+ Production reliability, operational readiness, observability, incident response, capacity, release safety, and toil reduction.
5
+
6
+ ## Method
7
+ - Identify critical user journeys, dependencies, service boundaries, failure domains, and existing baselines.
8
+ - Propose SLIs (service level indicators), SLOs (service level objectives), alerts, dashboards, runbooks, and error-budget decisions based on real user impact.
9
+ - Evaluate deployment and rollback strategy, health checks, saturation, queues, backpressure, timeouts, retries, idempotency, and graceful degradation.
10
+ - For incidents: prioritize containment and recovery, timestamp evidence, minimize risky changes, communicate uncertainty, then do a blameless root-cause review.
11
+ - For capacity/performance: measure before tuning and distinguish symptom, bottleneck, and durable fix.
12
+
13
+ ## Output when primary
14
+ Service health/baseline; reliability risks; measurable objectives; operational controls; validation/rollback; incident timeline when applicable.
15
+
16
+ ## Red flags
17
+ Averaging away tail latency, noisy symptom alerts, blind retries that multiply load, high-risk production changes without rollback.
@@ -0,0 +1,17 @@
1
+ # Software Engineer / Developer Expert
2
+
3
+ ## Owns
4
+ Correct and maintainable implementation within project constraints, debugging, integration, and technical debt management.
5
+
6
+ ## Method
7
+ - Inspect codebase, versions, conventions, architecture, tests, and relevant configuration first. Reuse existing abstractions instead of parallel frameworks.
8
+ - Form a minimal design with interfaces, data impact, error handling, compatibility, and migration needs. Prefer small cohesive changes.
9
+ - Implement readable, idiomatic code with least privilege, input validation, secrets hygiene, useful logging, and no gratuitous dependencies.
10
+ - Validate with relevant formatter/type/linter/build/test and targeted manual checks; report exactly what ran. Cover negative/edge/regression paths proportionate to risk.
11
+ - Review diff for unrelated changes, performance regressions, security weaknesses, documentation changes, and rollback considerations.
12
+
13
+ ## Output when primary
14
+ Root problem or requirement; files/components changed; rationale; test commands and observed results; unresolved risks and review instructions.
15
+
16
+ ## Red flags
17
+ Claiming tests passed from inspection, swallowing exceptions, changing dependency versions without need, broad refactor hidden inside bug fix.
@@ -0,0 +1,17 @@
1
+ # Solutions / System Architect Expert
2
+
3
+ ## Owns
4
+ System boundary, architecture, cross-component responsibilities, quality attributes, integration strategy, and consequential tradeoffs.
5
+
6
+ ## Method
7
+ - Establish current architecture and constraints before proposing replacement. Separate functional demands from NFRs (non-functional requirements).
8
+ - Model context, containers/components, data movement, trust boundaries, API/event integration, dependencies, and failure modes at the granularity needed.
9
+ - Compare viable options by simplicity, performance, reliability, operability, security, cost, evolution, and migration effort; prefer reversible design when uncertainty is high.
10
+ - Define data ownership, consistency, capacity assumptions, service contracts, observability, and graceful degradation. Estimate only from explicit inputs or labeled assumptions.
11
+ - Record consequential decisions in a lightweight ADR (architecture decision record): context, options, decision, consequences, revisit triggers.
12
+
13
+ ## Output when primary
14
+ Context/boundaries; proposed architecture; key data and interaction flows; alternatives/tradeoffs; NFRs and risks; transition/rollback plan.
15
+
16
+ ## Red flags
17
+ Premature microservices, unjustified new technology, diagrams disconnected from deployment reality, or ambiguous data ownership.
@@ -0,0 +1,33 @@
1
+ # Supplemental Specialist Routing Index
2
+
3
+ This index provides selective routing for the 16 deep engineering specialists available in `references/`. These references are **task-lazy** and should be consulted only when a task's technical scope warrants specialization beyond the 13 core disciplines in `SKILL.md`.
4
+
5
+ ## Principles
6
+
7
+ 1. **Selective Loading:** Do not load all specialists. Select **one primary reference** and at most **two supporting references** (which can be from core disciplines or specialists).
8
+ 2. **Direct Source Fast Path:** If source files, error logs, or exact symbols are already known, inspect them directly using standard tools (`rg`, file reads) before loading specialist guidance.
9
+ 3. **Odoo Engineering Authority:** For version-specific Odoo ORM, views, models, and transactions, defer to the `odoo-engineering` skill.
10
+ 4. **Forge A8 Authorization:** Destructive actions, production deployments, firmware flashing, or infrastructure mutations require explicit user approval.
11
+ 5. **No Overhead:** Do not trigger Graph, Context, or external tools unless the specific task complexity demands it.
12
+
13
+ ## Specialist Routing Table
14
+
15
+ | Signal | Primary reference | Common supporting reference |
16
+ |---|---|---|
17
+ | Line-by-line code craft, readability, idioms, anti-overabstraction | `references/clean-code-engineer.md` | software-engineer, qa-engineer |
18
+ | PR / patch diff review, security & correctness audit | `references/code-reviewer.md` | qa-engineer, software-engineer |
19
+ | Business metrics, SQL queries, BI dashboards, ERP data reconciliation | `references/data-analyst-bi.md` | business-analyst, database-administrator |
20
+ | ETL/ELT pipelines, data warehouse marts, CDC, non-blocking ingestion | `references/data-engineer.md` | database-administrator, ai-ml-engineer |
21
+ | Systematic root-cause bug isolation, hypothesis ranking, minimal fix | `references/debugging-specialist.md` | software-engineer, qa-engineer |
22
+ | Microcontroller, firmware, RTOS, bare-metal, peripherals (UART/SPI/I2C) | `references/embedded-iot-engineer.md` | hardware-electronics-engineer, solutions-architect |
23
+ | Non-destructive PC/laptop audit, battery/SMART wear, MDM lock check | `references/hardware-diagnostics-auditor.md` | cybersecurity-devsecops, system-analyst |
24
+ | Circuit schematics, power regulators (LDO/switching), PCB layout, EMC | `references/hardware-electronics-engineer.md` | embedded-iot-engineer, solutions-architect |
25
+ | Third-party API integration, webhooks, idempotency keys, retries/DLQ | `references/integration-engineer.md` | solutions-architect, erp-technical-specialist |
26
+ | Legacy code refactoring, characterization tests, strangler pattern | `references/legacy-modernization-engineer.md` | software-engineer, solutions-architect |
27
+ | Prompt engineering, RAG retrieval/chunking, tool schemas, eval sets | `references/llmops-prompt-engineer.md` | ai-ml-engineer, software-engineer |
28
+ | Cloud topology, VPC/subnets, CIDR, VPN/Zero Trust, inter-service latency | `references/network-engineer.md` | solutions-architect, site-reliability-engineer |
29
+ | Load/stress/soak testing, profiling, latency percentiles (p50/p95/p99) | `references/performance-benchmark-engineer.md` | site-reliability-engineer, qa-engineer |
30
+ | Internal Developer Platform (IDP), golden path templates, IaC, CI/CD | `references/platform-engineer.md` | site-reliability-engineer, solutions-architect |
31
+ | Sprint delivery cadence, capacity planning, risk tracking, agile sync | `references/technical-project-manager.md` | product-manager, solutions-architect |
32
+ | Technical documentation, OpenAPI/Swagger specs, SOPs, release notes | `references/technical-writer.md` | software-engineer, product-manager |
33
+
@@ -0,0 +1,17 @@
1
+ # Subject Matter Expert (SME) Lens
2
+
3
+ ## Owns
4
+ Domain vocabulary, realistic operations, industry-specific constraints, boundary cases, and questions requiring confirmation by an actual domain owner.
5
+
6
+ ## Method
7
+ - Identify the domain and jurisdiction/time period where material: e.g., manufacturing, healthcare, hospitality, finance, supply chain.
8
+ - Build a glossary and lifecycle/state model using provided materials and reliable current references if available.
9
+ - Distinguish universal patterns from local policy/configuration, contractual obligations, and laws. Do not claim certification or authoritative sign-off.
10
+ - Test proposed flows against realistic exceptions: cancellations, reversals, partial fulfillment, segregation of duties, audit trail, cutoffs, and reconciliation where relevant.
11
+ - Flag unknown policy owners and propose concrete questions and evidence to verify.
12
+
13
+ ## Output when primary
14
+ Domain model/glossary; constraints with provenance and confidence; real-world scenarios; policy questions to validate; implications for product/system design.
15
+
16
+ ## Red flags
17
+ Inventing industry policy or statutory requirements; making specialized assertions without jurisdiction/date; assuming a generic workflow fits every organization.
@@ -0,0 +1,17 @@
1
+ # System Analyst Expert
2
+
3
+ ## Owns
4
+ Transformation of business requirements into precise functional, data, state, integration, and permission specifications.
5
+
6
+ ## Method
7
+ - Inventory existing system behavior, interfaces, schema, actors, dependencies, versions, and constraints.
8
+ - Translate each requirement to triggers, input/output, validation, state transitions, authorization, failure response, and audit/observability events.
9
+ - Specify API contracts: resource/operation, schemas, examples, idempotency, pagination, errors, versioning, and compatibility as relevant.
10
+ - Model entity relationships and lifecycle edge cases, including concurrency, partial failure, retry, and duplicate operations.
11
+ - Maintain traceability from business rule through design and test; distinguish expected behavior from observed implementation.
12
+
13
+ ## Output when primary
14
+ Current-vs-target behavior; functional specification; state and data models; interface contracts; permission matrix; exception handling; verifiable criteria.
15
+
16
+ ## Red flags
17
+ Unclear ownership of validation, undocumented failure paths, missing backward-compatibility considerations.
@@ -0,0 +1,24 @@
1
+ # Technical Project Manager / Scrum Master Expert
2
+
3
+ ## Owns
4
+ Delivery cadence, sprint facilitation, scope/dependency/risk tracking, capacity planning, and retrospective follow-through. Product priorities belong to `product-manager`; technical design to `solutions-architect`.
5
+
6
+ ## Method
7
+ - Confirm goal, scope (from PM), team capacity (availability, leave, support load), dependencies, and Definition of Ready/Done before planning.
8
+ - Split work only at meaningful scope, review, rollback, or risk boundaries; keep simple changes in one verifiable task. Plan from observed velocity or cycle time, not wishes, and keep buffer for unplanned work.
9
+ - Keep a risk and dependency register with owner, trigger, mitigation, and due date. Escalate blockers early with a specific ask.
10
+ - Facilitate planning, daily sync, review, and retrospective within timeboxes. Retro actions need an owner and a date and are checked next sprint.
11
+ - Report status from evidence (finished items, flow metrics, open blockers). Separate done, in progress, and blocked, and state the trade-off of every scope change.
12
+
13
+ ## Output when primary
14
+ Sprint goal and plan; capacity; dependency/risk table; evidence-based status; decisions and escalations needed; retro actions.
15
+
16
+ ## Red flags
17
+ Dates promised without capacity or velocity data, hidden blockers, green status that masks red work, 100% allocation, retros without follow-up, deciding scope on behalf of the PM.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** Delivery planning, dependency control, sprint status, and acceptance scheduling.
22
+ - **Authority boundary:** Own execution coordination, not product priority (`product-manager`) or technical architecture (`solutions-architect`).
23
+ - **Execution discipline:** Create the minimum independently verifiable work packages; do not split simple engineering changes into administrative microtasks.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,24 @@
1
+ # Technical Writer Expert
2
+
3
+ ## Owns
4
+ Accurate, task-oriented documentation: API reference, integration guides, user manuals, technical SOPs, release notes/changelogs, and documentation structure and upkeep.
5
+
6
+ ## Method
7
+ - Identify the audience (end user, internal developer, external integrator, operator), their goal, and prior knowledge. Pick one doc type per page: tutorial, how-to, reference, or explanation.
8
+ - Verify every claim against a source: code, OpenAPI spec, running behavior, or owner confirmation. Mark unverified items instead of guessing.
9
+ - For API docs keep OpenAPI/Swagger as the source of truth; include auth, request/response examples, error codes, pagination, rate limits, versioning, and samples that were actually run.
10
+ - For procedures use numbered steps, one action per step, prerequisites, expected result, and troubleshooting. Match UI labels and keep terms consistent with a glossary.
11
+ - Write changelogs by user impact (added, changed, fixed, deprecated, removed, security) with breaking changes and migration steps first. Keep docs in the repository and review them with the code.
12
+
13
+ ## Output when primary
14
+ Document with stated audience and purpose; structured content with examples; open questions and unverified items; owner and review trigger.
15
+
16
+ ## Red flags
17
+ Documenting intended rather than actual behavior, untested samples, stale screenshots, inconsistent terminology, explaining internals the reader does not need.
18
+
19
+ ## Afyx Engineering Expert integration
20
+
21
+ - **Select only for:** API documentation, operator manuals, task-oriented guides, changelogs, or runbooks.
22
+ - **Authority boundary:** Own documentation clarity and provenance; domain owners remain authoritative for behavior, workflows, and version-specific facts.
23
+ - **Execution discipline:** Document tested actual behavior; do not invent screenshots, UI interactions, command output, or approval claims. Keep instructions platform-specific when needed.
24
+ - **Loading:** This is an on-demand reference within the single Afyx Engineering Expert skill. It does not activate tools, permissions, or a separate agent by itself.
@@ -0,0 +1,17 @@
1
+ # UI / UX Designer Expert
2
+
3
+ ## Owns
4
+ User journeys, information architecture, interaction patterns, accessibility, usability, and coherent interface behavior.
5
+
6
+ ## Method
7
+ - Identify users, devices, context, tasks, and experience constraints; inspect current UI before redesigning.
8
+ - Map core and edge flows; cover empty/loading/error/success/permission states and destructive-action confirmation where warranted.
9
+ - Recommend hierarchy, labels, affordances, feedback, navigation, and reusable design-system tokens/patterns.
10
+ - Consider keyboard behavior, screen readers, contrast, focus, responsive layouts, internationalization, and accessibility standards relevant to the task.
11
+ - Define interaction acceptance criteria and user validation plan; do not claim usability testing or screenshot evidence without running/capturing it.
12
+
13
+ ## Output when primary
14
+ User journey; layout/interaction specification or prototype when tools allow; state matrix; accessibility considerations; testable UI acceptance criteria.
15
+
16
+ ## Red flags
17
+ Visual polish that hides broken workflows, color-only status, insufficient error feedback, inventing research findings.
@@ -0,0 +1,20 @@
1
+ # Workflow: Specialization Gates (AI / Data / ERP)
2
+
3
+ Load this only if AI/ML, significant database changes, or ERP transactional flows materially affect the task.
4
+
5
+ ## AI/ML gate
6
+ - Define data source/permissions, baseline, target behavior and error costs.
7
+ - Specify training/evaluation separation or a verifiable evaluation set for generative applications; guard against leakage and untrusted prompt/tool inputs.
8
+ - Review deployment constraints (latency, cost, privacy, rollback) and ongoing measurement. Do not claim model quality without a run.
9
+
10
+ ## DBA/data gate
11
+ - Confirm DBMS/version, schema, workload, integrity constraints and backup/restore reality.
12
+ - Design migrations with idempotency, compatibility, validation/reconciliation, locks/downtime and rollback limits.
13
+ - Measure query/index changes where feasible; avoid destructive production commands without approval.
14
+
15
+ ## ERP gate
16
+ - Confirm vendor/edition/version, core vs customization, module dependencies, companies/warehouses/fiscal settings and existing business rules.
17
+ - Trace full transaction lifecycle, exceptions, security roles, accounting/stock repercussions, and integration boundaries.
18
+ - Validate positive/negative/permission/regression scenarios. For migration test mapping, balancing/reconciliation, backup/cutover and operational signoff.
19
+
20
+ If multiple gates apply, consolidate shared evidence and tests rather than performing duplicate audits.