@exaudeus/workrail 0.3.0 → 0.4.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@exaudeus/workrail",
-  "version": "0.3.0",
+  "version": "0.4.1",
   "description": "MCP server for structured workflow orchestration and step-by-step task guidance",
   "license": "MIT",
   "bin": {

package/workflows/mr-review-workflow.json

@@ -30,6 +30,8 @@
     "Document I/O: create/update the file at reviewDocPath using file-writing tools. If the agent cannot write files, output the exact doc content with the intended path so the user can save it.",
     "If a function requires inputs that are missing, state exactly what is missing and pause. Otherwise, do not ask for confirmation.",
     "If prompt text and a function definition conflict, prefer the function definition for the how, and the step prompt for the what.",
+    "Phase 1a is user-driven: request ticket/dev/design/BRD; wait for user reply; do not search or analyze code in this step.",
+    "Do not self-confirm. Only proceed after explicit user confirmation or 'skip'; then set contextGatheringComplete=true or skipContextGathering=true.",
     "discoverModulePatterns(): Compute moduleRoot; scan only moduleRoot to build patternCatalog.",
     "discoverModulePatterns() details: moduleRoot = nearest common ancestor, clamped to package/src; catalog = naming, DI, errors, logging, tests, layout.",
     "consolidatePatternFindings(): Compare changes to patternCatalog (scoped to moduleRoot).",
@@ -85,9 +87,13 @@
     {
       "id": "phase-1a-llm-context-gathering",
       "title": "Phase 1a: Comprehensive Context Gathering",
-      "prompt": "To perform a thorough review, I need more than just the code diff. Please provide the following:\n\n1. **Ticket(s):** Link(s) and, if possible, paste the full text (requirements, acceptance criteria).\n2. **Dev/Design Docs:** Dev plan; design/architecture docs; and any BRD or decision records relevant to this MR.\n3. **Technical Context:** If this change relies on other parts of the codebase, provide relevant code snippets or file contents; include coding standards or module conventions if available.\n\nI will synthesize this information to build a comprehensive context for my review.\n\n**Confirmation & Gating:**\n- After you confirm that I have the right context (or explicitly instruct me to skip), I will set `contextGatheringComplete = true` or `skipContextGathering = true` accordingly.\n- I will not proceed to deeper analysis until one of these is set.",
+      "prompt": "To perform a thorough review, I need you to provide source documents. In this step I will not search or analyze code—I'm waiting for your inputs. Please provide:\n\n1. **Ticket(s):** Link(s) and, if possible, paste the full text (requirements, acceptance criteria).\n2. **Dev/Design Docs:** Dev plan; design/architecture docs; any BRD/ADR/decision records relevant to this MR.\n3. **Technical Context:** If this change relies on other parts of the codebase, paste relevant code snippets or file contents; include coding standards or module conventions if available.\n\nI will synthesize this information once received.\n\n**Confirmation & Gating:**\n- After you confirm that I have the right context (or explicitly instruct me to skip), I will set `contextGatheringComplete = true` or `skipContextGathering = true` accordingly.\n- I will not proceed to deeper analysis until one of these is set.",
       "agentRole": "You are a context-aware analyst ensuring you have all necessary information before starting a deep review.",
-      "guidance": ["A high-quality review depends on high-quality context. The more information provided here, the more accurate the review will be."],
+      "guidance": [
+        "User-driven step: ask for ticket, dev plan, design/architecture docs, BRD/ADR, and any coding standards.",
+        "Do not search or analyze code in this step; wait for user inputs.",
+        "Only after explicit confirmation or 'skip' set contextGatheringComplete=true or skipContextGathering=true."
+      ],
       "runCondition": { "var": "complexity", "not_equals": "Trivial" },
       "requireConfirmation": true
     },
@@ -101,9 +107,10 @@
         "Scope strictly to moduleRoot and changed files plus immediate neighbors.",
         "Cap per-file neighbors and graph size to avoid noise; prefer edges directly related to the diff.",
         "Write full details to the live doc; in chat, keep output minimal.",
-        "Render 'ASCII Call Graphs' under 'Bounded Call Graph': edges 'A -> B', width ≤100, highlight changed as [*name*], include tiny legend.",
-        "Paths: for each changed symbol show top 2–3 inbound and outbound paths, depth ≤2, dedupe, shortest-first then alphabetical.",
-        "Caps: ≤12 nodes, ≤20 edges across graphs; if exceeded, fall back to an adjacency list with per-edge counts.",
+        "Small Multiples: for each changed symbol, render mini path lines (≤6 nodes) for top 1–2 inbound and 1–2 outbound paths; width ≤100; highlight changed as [*name*]; tag HOT where applicable.",
+        "Path Lists with Aliasing: collapse repeated subpaths using deterministic aliases (A1, A2, ...); reuse aliases across paths; append an 'Alias Legend' listing alias → full subpath.",
+        "Ordering: HOT-first, then shortest path, then alphabetical; depth ≤2; dedupe identical paths.",
+        "Caps: ≤5 anchors; ≤8 total paths; ≤12 nodes across all small multiples; if exceeded, render 'Adjacency Summary' with per-edge counts instead.",
         "Flow Anchors: map entrypoints to changed symbols; table 'Anchor | Type | Target Symbol | Notes'; cap ≤5 anchors.",
         "Hot-Path ranking: score edges (diff touch, anchor, ticket keyword, test ref); tag HOT if score ≥ threshold; prefer HOT paths in ASCII output.",
         "Legend in doc: [*name*]=changed; HOT=high-impact; 'A -> B'=calls; 'Route -> Handler'; 'Job -> Worker'.",
@@ -170,7 +177,7 @@
           "guidance": [
             "Targets: naming clarity; small readability/style inconsistencies; redundant comments; commented/dead code; import/order polish; safe micro-perf; test naming/structure.",
             "Quotas: if majorIssuesFound==true aim 5–8 (cap 12); else aim 8–12 (cap 15); ≤3 per category; 1–2 lines each; dedupe; group by file; prefix 'Nit:'.",
-            "Add ready-to-copy MR comments for the top 3 nits; log the revision.",
+            "MUST add ready-to-copy MR comments for the top 3 nits (file:line, short title, 1–2 line rationale, optional tiny diff); log the revision.",
             "Append Nit grouped by file: 'path:line — nit text'."
           ],
           "runCondition": { "var": "analysisDepth", "equals": 4 },
@@ -238,10 +245,11 @@
         "Dedupe: identical path:line + similar text → keep one; if overlap with Hygiene/Patterns, keep in the most specific section and cross-reference in the other.",
         "Caps: ≤15 total; ≤3 per category; if exceeded, append a single 'More items omitted due to caps.' line.",
         "Cross-link any existing MR comments to their nit rows (e.g., 'MR: MR-##').",
+        "If fewer than 3 Nit MR comments exist overall, create minimal ready-to-copy MR comments for the highest-signal remaining nits to reach 3 (file:line, short title, 1–2 line rationale, optional tiny diff).",
         "Chat: minimal recap with doc path and counts; full details live in the doc."
       ],
       "runCondition": { "and": [ { "var": "complexity", "not_equals": "Trivial" }, { "or": [ { "var": "contextGatheringComplete", "equals": true }, { "var": "skipContextGathering", "equals": true } ] } ] },
-      "functionReferences": ["upsertSection()", "logRevision()"],
+      "functionReferences": ["appendMRComment()", "upsertSection()", "logRevision()"],
       "requireConfirmation": false
     },
     {
@@ -257,6 +265,7 @@
         "Risk Heatmap table: rows 'Risk | Area | Likelihood(L/M/H) | Impact(L/M/H) | Mitigation(≤1 line)'; sort by Impact desc then Likelihood desc; cap ≤6 rows; add 'omitted due to caps' line if needed.",
         "Dedupe patches and MR comments; ensure every Critical/Major has either a patch or a copy-ready MR comment.",
         "If any Critical/Major lacks an MR comment, generate a minimal ready-to-copy comment (file:line, short title, 1–2 line rationale, optional tiny diff snippet).",
+        "Ensure at least 3 Nit MR comments exist overall; if fewer, create minimal Nit MR comments for the top remaining nits to reach 3.",
         "Group MR comments by file and dedupe similar entries; add a 'Missing Comments Summary' note if any comments were auto-added.",
         "Cross-link each MR comment to its related Proposed Patch when applicable ('See Proposed Patch: <title>').",
         "Keep all caps deterministic; if any list exceeds its cap, add a single trailing 'More items omitted due to caps.' line."