@rune-kit/rune 2.11.0 → 2.12.2

package/skills/session-bridge/scripts/load-invariants.js

@@ -0,0 +1,397 @@
+#!/usr/bin/env node
+
+/**
+ * load-invariants.js — Parse `.rune/INVARIANTS.md` at session start.
+ *
+ * Responsibilities:
+ *   - Read the file if it exists (silent no-op otherwise).
+ *   - Skip the `## Archived` section entirely (rules there are intentionally
+ *     retired; including them would re-activate historical noise).
+ *   - Extract active rules with WHAT / WHERE / WHY + section label.
+ *   - Produce a token-budgeted preview string (≤ ~500 tokens by default) that
+ *     session-bridge injects verbatim into the agent's session-start summary.
+ *   - Detect staleness: if mtime > 30 days old, flag so session-bridge can warn
+ *     once per session.
+ *
+ * Returned shape:
+ *   {
+ *     loaded: boolean,
+ *     path: string,
+ *     stale: boolean,           // mtime > 30 days old
+ *     stats: { danger, critical, state, cross, archivedSkipped, total },
+ *     rules: Rule[],            // [{ section, title, what, where: string[], why }]
+ *     preview: string,          // agent-visible bullet list, capped by budget
+ *     overflow: number,         // rules present but NOT in preview (budget overflow)
+ *   }
+ *
+ * Usage as library:
+ *   import {
+ *     loadInvariants,
+ *     matchesInvariant,
+ *     findMatchingInvariants,
+ *   } from './load-invariants.js';
+ *
+ *   const result = await loadInvariants({ root, budgetTokens: 500 });
+ *   // Blast-radius check (used by logic-guardian, Pro autopilot Step 0.5):
+ *   const hits = findMatchingInvariants('skills/cook/foo.js', result.rules);
+ *
+ * Usage as CLI:
+ *   node load-invariants.js --root <project-root> [--json]
+ */
+
+import { existsSync } from 'node:fs';
+import { readFile, stat } from 'node:fs/promises';
+import path from 'node:path';
+import { parseArgs } from 'node:util';
+
+const DEFAULT_BUDGET_TOKENS = 500;
+const STALE_AGE_DAYS = 30;
+
+/**
+ * Rough token estimator. Uses Unicode scalar count (`[...str].length`) rather
+ * than `str.length`, because:
+ *   - `str.length` in JS = UTF-16 code units → emoji outside the BMP count as 2
+ *     (e.g. `🔒`.length === 2) even though tokenizers treat them as ~1 token.
+ *   - CJK / Vietnamese text tokenizes closer to 1 char ≈ 1 token, not 1/4 —
+ *     scalar count gives a safer lower-bound denominator.
+ * Not exact tokenization — just a deterministic ceiling that doesn't drift
+ * wildly on non-ASCII content.
+ */
+function estimateTokens(str) {
+  let scalars = 0;
+  for (const _ of str) scalars += 1;
+  return Math.ceil(scalars / 3);
+}
+
+/**
+ * Walk `text` line-by-line, tracking triple-backtick / triple-tilde fence
+ * state. Yields `{ line, inFence, offset }` for each line. Shared by
+ * `stripArchived` and `parseInvariants` so markdown constructs inside fenced
+ * code blocks (e.g. a `## Archived` line inside a ```markdown example) never
+ * trigger false-positive section breaks or rule headers.
+ */
+function* walkLines(text) {
+  let offset = 0;
+  let inFence = false;
+  let fenceDelim = null;
+  while (offset <= text.length) {
+    const eol = text.indexOf('\n', offset);
+    const lineEnd = eol === -1 ? text.length : eol;
+    const line = text.slice(offset, lineEnd);
+    const trimmed = line.trimStart();
+    if (inFence) {
+      if (trimmed.startsWith(fenceDelim)) {
+        inFence = false;
+        fenceDelim = null;
+      }
+      yield { line, inFence: true, offset };
+    } else {
+      if (trimmed.startsWith('```') || trimmed.startsWith('~~~')) {
+        inFence = true;
+        fenceDelim = trimmed.startsWith('```') ? '```' : '~~~';
+        yield { line, inFence: true, offset };
+      } else {
+        yield { line, inFence: false, offset };
+      }
+    }
+    if (eol === -1) break;
+    offset = eol + 1;
+  }
+}
+
+/**
+ * Strip the `## Archived` section (and anything below it) from `text`.
+ * Archived rules are intentionally retired — loading them would re-surface
+ * historical noise. Fence-aware: `## Archived` inside a code block does NOT
+ * terminate active rules.
+ */
+export function stripArchived(text) {
+  for (const { line, inFence, offset } of walkLines(text)) {
+    if (inFence) continue;
+    if (/^## Archived\s*$/.test(line)) {
+      return text.slice(0, offset);
+    }
+  }
+  return text;
+}
+
+/**
+ * Parse an INVARIANTS.md body (with archived already stripped) into rule
+ * objects. Walks `### SectionTitle` and `#### RuleTitle` blocks; expects each
+ * rule block to carry `- **WHAT**:`, `- **WHERE**:`, `- **WHY**:` lines.
+ * Missing lines are tolerated — they become empty strings / empty arrays.
+ */
+export function parseInvariants(text) {
+  const rules = [];
+  const SECTION_RE = /^##\s+(Danger Zones|Critical Invariants|State Machine Rules|Cross-File Consistency)\b/;
+  const SUBSECTION_RE = /^###\s+(Danger Zones|Critical Invariants|State Machine Rules|Cross-File Consistency)\b/;
+  const RULE_HEADER_RE = /^####\s+(.+?)\s*$/;
+  const FIELD_RE = /^-\s+\*\*(WHAT|WHERE|WHY)\*\*:\s*(.+)$/;
+
+  let currentSection = null;
+  let current = null;
+
+  const flush = () => {
+    if (current && current.title) rules.push(current);
+    current = null;
+  };
+
+  for (const { line: raw, inFence } of walkLines(text)) {
+    // Fenced code blocks can legitimately contain text that looks like
+    // section headers or field lines (e.g. a markdown example inside a WHY).
+    // Ignore them while inside a fence.
+    if (inFence) continue;
+    const line = raw.trimEnd();
+    const sectionMatch = SECTION_RE.exec(line) || SUBSECTION_RE.exec(line);
+    if (sectionMatch) {
+      flush();
+      currentSection = sectionLabelToKey(sectionMatch[1]);
+      continue;
+    }
+    const ruleMatch = RULE_HEADER_RE.exec(line);
+    if (ruleMatch && currentSection) {
+      flush();
+      current = {
+        section: currentSection,
+        title: ruleMatch[1].trim(),
+        what: '',
+        where: [],
+        why: '',
+      };
+      continue;
+    }
+    if (!current) continue;
+    const fieldMatch = FIELD_RE.exec(line);
+    if (!fieldMatch) continue;
+    const [, key, value] = fieldMatch;
+    if (key === 'WHAT') current.what = value.trim();
+    else if (key === 'WHY') current.why = value.trim();
+    else if (key === 'WHERE') current.where = extractGlobs(value);
+  }
+  flush();
+  return rules;
+}
+
+function sectionLabelToKey(label) {
+  switch (label) {
+    case 'Danger Zones':
+      return 'danger';
+    case 'Critical Invariants':
+      return 'critical';
+    case 'State Machine Rules':
+      return 'state';
+    case 'Cross-File Consistency':
+      return 'cross';
+    default:
+      return 'other';
+  }
+}
+
+/**
+ * Convert a minimal glob pattern to a RegExp. Supports the subset of globs
+ * seen in INVARIANTS.md `WHERE` entries:
+ *   - `**` matches any number of path segments (including zero)
+ *   - `*` matches anything except `/`
+ *   - `?` matches a single non-`/` character
+ *   - All other regex metacharacters are escaped literally
+ * Paths are normalized to forward slashes before matching.
+ */
+function globToRegExp(glob) {
+  const normalized = glob.replace(/\\/g, '/');
+  let re = '';
+  let i = 0;
+  while (i < normalized.length) {
+    const c = normalized[i];
+    if (c === '*' && normalized[i + 1] === '*') {
+      // `**/` = zero or more segments
+      if (normalized[i + 2] === '/') {
+        re += '(?:.*/)?';
+        i += 3;
+      } else {
+        re += '.*';
+        i += 2;
+      }
+    } else if (c === '*') {
+      re += '[^/]*';
+      i += 1;
+    } else if (c === '?') {
+      re += '[^/]';
+      i += 1;
+    } else if ('.+^$()[]{}|\\'.includes(c)) {
+      re += `\\${c}`;
+      i += 1;
+    } else {
+      re += c;
+      i += 1;
+    }
+  }
+  return new RegExp(`^${re}$`);
+}
+
+/**
+ * Return true if `filePath` matches any glob in a rule's `where[]`.
+ * Normalizes paths to forward slashes first so Windows and POSIX callers get
+ * identical results. Consumers (logic-guardian pre-edit gate, autopilot
+ * pre-flight gate) use this to decide whether a planned edit touches an
+ * invariant-protected path.
+ */
+export function matchesInvariant(filePath, rule) {
+  if (!rule || !Array.isArray(rule.where) || rule.where.length === 0) return false;
+  const normalized = filePath.replace(/\\/g, '/').replace(/^\.\//, '');
+  for (const glob of rule.where) {
+    if (globToRegExp(glob).test(normalized)) return true;
+  }
+  return false;
+}
+
+/**
+ * Return every rule whose WHERE globs match `filePath`. Order preserved from
+ * the loaded rules array. Use this when an edit blast radius needs to surface
+ * every active invariant that applies.
+ */
+export function findMatchingInvariants(filePath, rules) {
+  return rules.filter((rule) => matchesInvariant(filePath, rule));
+}
+
+function extractGlobs(raw) {
+  const BACKTICK_RE = /`([^`]+)`/g;
+  const globs = [];
+  let match;
+  BACKTICK_RE.lastIndex = 0;
+  while ((match = BACKTICK_RE.exec(raw))) {
+    globs.push(match[1].trim());
+  }
+  // Fallback: if no backticks, split on commas/semicolons.
+  if (globs.length === 0) {
+    return raw
+      .split(/[,;]/)
+      .map((s) => s.trim())
+      .filter(Boolean);
+  }
+  return globs;
+}
+
+/**
+ * Render an agent-visible preview. Caps at ~budgetTokens by dropping tail rules
+ * and appending an overflow marker. Priority order: danger → critical → state → cross.
+ */
+export function renderPreview(rules, { budgetTokens = DEFAULT_BUDGET_TOKENS } = {}) {
+  if (rules.length === 0) {
+    return { preview: '', overflow: 0 };
+  }
+  const priority = { danger: 0, critical: 1, state: 2, cross: 3, other: 4 };
+  const ordered = [...rules].sort((a, b) => (priority[a.section] ?? 9) - (priority[b.section] ?? 9));
+
+  const ICON = { danger: '⚠', critical: '🔒', state: '🔁', cross: '🔗', other: '•' };
+  const header = '📎 Active Invariants (.rune/INVARIANTS.md)';
+  const lines = [header];
+  let tokens = estimateTokens(header);
+  let shown = 0;
+
+  for (const rule of ordered) {
+    const icon = ICON[rule.section] ?? '•';
+    const wherePreview = rule.where.length > 0 ? rule.where.slice(0, 2).join(', ') : rule.title;
+    const line = `${icon}  ${wherePreview} — ${rule.what || rule.title}`;
+    const cost = estimateTokens(line);
+    if (tokens + cost > budgetTokens) break;
+    lines.push(line);
+    tokens += cost;
+    shown += 1;
+  }
+
+  const overflow = ordered.length - shown;
+  if (overflow > 0) {
+    lines.push(`…+${overflow} more rule${overflow === 1 ? '' : 's'} in .rune/INVARIANTS.md`);
+  }
+  return { preview: lines.join('\n'), overflow };
+}
+
+export async function loadInvariants({ root, budgetTokens = DEFAULT_BUDGET_TOKENS } = {}) {
+  if (!root) throw new Error('loadInvariants: root is required');
+
+  const invariantsPath = path.join(root, '.rune', 'INVARIANTS.md');
+  const empty = {
+    loaded: false,
+    count: 0,
+    path: invariantsPath,
+    stale: false,
+    stats: { danger: 0, critical: 0, state: 0, cross: 0, total: 0, archivedSkipped: false },
+    rules: [],
+    preview: '',
+    overflow: 0,
+  };
+  if (!existsSync(invariantsPath)) return empty;
+
+  let raw;
+  try {
+    raw = await readFile(invariantsPath, 'utf8');
+  } catch {
+    return empty;
+  }
+
+  const beforeArchive = stripArchived(raw);
+  const archivedSkipped = beforeArchive.length !== raw.length;
+  const rules = parseInvariants(beforeArchive);
+
+  const stats = {
+    danger: rules.filter((r) => r.section === 'danger').length,
+    critical: rules.filter((r) => r.section === 'critical').length,
+    state: rules.filter((r) => r.section === 'state').length,
+    cross: rules.filter((r) => r.section === 'cross').length,
+    total: rules.length,
+    archivedSkipped,
+  };
+
+  let stale = false;
+  try {
+    const st = await stat(invariantsPath);
+    const ageDays = (Date.now() - st.mtimeMs) / (1000 * 60 * 60 * 24);
+    stale = ageDays > STALE_AGE_DAYS;
+  } catch {
+    /* stat failure is non-fatal */
+  }
+
+  const { preview, overflow } = renderPreview(rules, { budgetTokens });
+
+  return {
+    loaded: rules.length > 0,
+    count: rules.length,
+    path: invariantsPath,
+    stale,
+    stats,
+    rules,
+    preview,
+    overflow,
+  };
+}
+
+async function main() {
+  const { values } = parseArgs({
+    options: {
+      root: { type: 'string' },
+      json: { type: 'boolean', default: false },
+      budget: { type: 'string', default: String(DEFAULT_BUDGET_TOKENS) },
+    },
+  });
+  const root = values.root ?? process.cwd();
+  const parsedBudget = Number.parseInt(values.budget, 10);
+  const budgetTokens = Number.isFinite(parsedBudget) && parsedBudget > 0 ? parsedBudget : DEFAULT_BUDGET_TOKENS;
+  const result = await loadInvariants({ root, budgetTokens });
+  if (values.json) {
+    process.stdout.write(`${JSON.stringify(result, null, 2)}\n`);
+    return;
+  }
+  if (!result.loaded) {
+    process.stdout.write('No invariants loaded — .rune/INVARIANTS.md missing or empty.\n');
+    return;
+  }
+  process.stdout.write(`${result.preview}\n`);
+  if (result.stale) process.stdout.write('\n⚠ Invariants file is stale (>30 days). Run `rune onboard --refresh`.\n');
+}
+
+if (import.meta.url === `file://${process.argv[1]}` || process.argv[1]?.endsWith('load-invariants.js')) {
+  main().catch((err) => {
+    process.stderr.write(`load-invariants: ${err.message}\n`);
+    process.exit(1);
+  });
+}

package/skills/slides/SKILL.md

@@ -140,3 +140,22 @@ Slide body content
 3. DO NOT require Marp installation — output is standard markdown that Marp can consume
 4. Keep slide count reasonable (5-15 for demos, 10-25 for talks)
 5. Always include speaker notes for non-trivial slides
+
+## Returns
+
+| Artifact | Format | Location |
+|----------|--------|----------|
+| Slide schema | JSON | `slides.json` (temporary) |
+| Presentation deck | Marp Markdown | `deck.md` or user-specified path |
+
+## Done When
+
+- Presentation type identified (demo/sprint/talk/tutorial/pitch)
+- JSON schema generated with correct slide types
+- build-deck.js executed (or fallback markdown generated)
+- Output file path presented to user
+- Preview/export commands provided
+
+## Cost Profile
+
+~500-1500 tokens input, ~300-800 tokens output. Sonnet for quality copy and structure.

package/skills/team/SKILL.md

@@ -5,7 +5,7 @@ context: fork
 agent: general-purpose
 metadata:
   author: runedev
-  version: "0.8.0"
+  version: "1.0.0"
   layer: L1
   model: opus
   group: orchestrator
@@ -105,6 +105,7 @@ Before decomposing, classify the task into a complexity tier. Each tier defines
 
 ## Called By (inbound)
 
+- `scaffold` (L1): decompose scaffolding into parallel workstreams
 - User: `/rune team <task>` direct invocation only
 
 ---

package/skills/test/SKILL.md

@@ -307,6 +307,12 @@ Save eval files as `skills/<name>/evals.md`. Each eval is a numbered scenario (E
 - `launch` (L1): pre-deployment test suite
 - `safeguard` (L2): writing characterization tests for legacy code
 - `review-intake` (L2): write tests for issues identified during review intake
+- `scaffold` (L1): generate initial test suite for new project
+- `graft` (L2): write integration tests for grafted code
+- `skill-forge` (L2): write tests for new skill functionality
+- `mcp-builder` (L2): write tests for MCP server tools
+- `debug` (L2): write regression test capturing the bug
+- `plan` (L2): reference test requirements in implementation plan
 
 ## Calls (outbound)
 

package/skills/verification/SKILL.md

@@ -204,6 +204,14 @@ None — pure runner using Bash for all checks. Does not invoke other skills.
 - `db` (L2): run migration in test environment
 - `perf` (L2): run benchmark scripts if configured
 - `skill-forge` (L2): verify newly created skill passes lint/type/build checks
+- `team` (L1): verify each parallel workstream before merge
+- `scaffold` (L1): verify scaffolded project builds and passes initial tests
+- `launch` (L1): pre-deploy verification gate
+- `mcp-builder` (L2): verify generated MCP server compiles and starts
+- `preflight` (L2): run verification as part of pre-commit quality gate
+- `logic-guardian` (L2): verify logic invariants hold after changes
+- `dependency-doctor` (L3): verify builds pass after dependency updates
+- `sast` (L3): run verification alongside static analysis
 
 ## Output Format