sinapse-ai 1.19.2 → 1.20.1

package/scripts/validate-agent-codenames.js

@@ -1,6 +1,6 @@
 #!/usr/bin/env node
 /**
- * validate-agent-codenames — AuditFinding AF-20260629 #32
+ * validate-agent-codenames — AuditFinding AF-20260629 #32 (+ AF-20260702 #1.10)
  *
  * Rejects DUPLICATE persona codenames (the `name:` field) shared across two or
  * more DISTINCT agents. A codename is an agent's public identity (Pixel, Litmus,
@@ -12,19 +12,36 @@
  *   - sinapse/agents/*.md             (master squad)
  *   - .sinapse-ai/development/agents/*.md (framework core agents)
  *
- * Codename extraction: the FIRST quoted `name: "X"` in the file. In every agent
- * format the persona block sits at the top, so the first quoted name is the
- * codename — knowledge-base `name:` entries (mental models, frameworks) appear
- * far below and are intentionally ignored. Files with no quoted `name:` (table /
- * frontmatter-only formats) are skipped — there is no codename to compare.
+ * Codename extraction (AF-20260702 #1.10 closes a blind spot here — the
+ * original version only recognized a QUOTED `name: "X"`, so an unquoted
+ * `name: X` (also valid YAML) was silently invisible to this guard):
+ *   1. YAML `name:` field, first occurrence, quoted OR unquoted
+ *      (`name: "Nexus"` and `name: Nexus` are both recognized). In every
+ *      agent format the persona block sits at the top, so the first match is
+ *      the codename — knowledge-base `name:` entries (mental models,
+ *      subagent types, command lists) are nested under a `- ` list marker or
+ *      appear far below and don't match the `^\s*name:` (no dash) anchor.
+ *   2. If no YAML `name:` is found, fall back to a PROSE declaration — the
+ *      two conventions actually in use across agent files with no YAML
+ *      persona block: `# Agent: {Name} — {Title}` and a markdown identity
+ *      bullet `- **Nome:** {Name}` / `- **Name:** {Name}`. This is how
+ *      design-orqx.md declares "Nexus" (plain markdown, not YAML) and why
+ *      the pre-fix guard could not see it at all.
+ *
+ * Files with neither form are skipped — there is no codename to compare.
  *
  * Allow-list: the master orchestrator (Imperator) is intentionally registered
  * under multiple entry-point files (snps-orqx / sinapse-orqx) — same logical
  * agent, dual register (Constitution Article XI). It is NOT a collision.
  *
+ * Known pending debt: KNOWN_COLLISIONS_PENDING_TRIAGE (below) lists real,
+ * pre-existing collisions this guard now detects but does NOT block on —
+ * they are out of scope for the story that fixed this guard's blind spots
+ * (see the constant's own comment) and are reported for visibility instead.
+ *
  * Exit codes:
- *   0  every codename is unique (modulo the allow-list)
- *   1  one or more codenames are shared by distinct agents
+ *   0  every codename is unique (modulo the allow-list and pending debt)
+ *   1  one or more NOT-yet-triaged codenames are shared by distinct agents
  *
  * Usage:
  *   node scripts/validate-agent-codenames.js
@@ -50,6 +67,51 @@ const SQUADS_DIR = path.join(PROJECT_ROOT, 'squads');
 // logical agent registered under multiple alias files (dual register).
 const ALLOWED_DUPLICATES = new Set(['Imperator']);
 
+// Known PRE-EXISTING collisions, NOT resolved by this guard (do not block).
+//
+// AF-20260702 #1.10 fixed two real blind spots: unquoted YAML `name:` and
+// prose-only persona declarations (e.g. `# Agent: {Name} — {Title}`, used
+// by several squads with no YAML persona block at all — squad-design's
+// dx-* roster in particular). Looking at every agent file with both
+// extraction paths surfaced 24 pre-existing codename collisions (Scope
+// through Vertex) that NO version of this guard had ever seen before.
+// Story onda1-s5 fixed one instance ("Nexus") and deliberately deferred
+// these 24 to a follow-up triage (Onda 2) rather than blanket-rename
+// without individual review (Constitution Article XI — Conservative
+// Default).
+//
+// Story onda2-p7 (2026-07-02) did that triage. Verdict for all 24: every
+// single one was a REAL collision between independently-named, established
+// agents (never a detector false-positive, never genuinely ambiguous) — the
+// palette of short evocative one-word codenames is small enough that 18+
+// squads picked the same words independently. Winner kept by: (1) core
+// framework agent beats squad agent, (2) squad orchestrator beats squad
+// specialist (both cited in .sinapse-ai/development/agents/snps-orqx.md's
+// master delegation-matrix), (3) among two specialists, semantic fit to the
+// role + which squad already yielded more collisions elsewhere (kept
+// squad-design's fragmented dx-* roster internally consistent rather than
+// stranding 2 of 8 with stale names). All 27 renames (3 codenames were
+// 3-way collisions) shipped with every live reference updated — own file,
+// squad README, task "**Agent:**" headers, cross-agent handoff mentions,
+// squad.yaml, and the master delegation-matrix in snps-orqx.md — verified
+// by re-running this guard (0 blocking, 0 pending) plus a dedicated
+// anchored-reference sweep (codename+old-agent-id co-occurrence, checked
+// across the WHOLE repo, not just the renamed agent's home squad). Full
+// old->new map and per-item reasoning: story `docs/stories/story-o2p7-agents-piloto.md`.
+//
+// This Set is intentionally EMPTY right now — every previously-known
+// pending collision was triaged and resolved. It stays wired into
+// `findCollisions()` as a NON-BLOCKING lane for whatever the guard's
+// extraction next surfaces (e.g. if squad-content's task format is ever
+// scanned the same way, or a new squad ships with an un-checked codename).
+// It is deliberately NOT the same mechanism as ALLOWED_DUPLICATES above
+// (that Set means "intentional, correct design"); an entry here means
+// "known bug, not yet fixed". Removing a name because it was fixed is
+// encouraged; ADDING a name to silence a *new* collision is not what this
+// list is for — triage it (rename, or prove it's a detector false-positive)
+// the same way this story did.
+const KNOWN_COLLISIONS_PENDING_TRIAGE = new Set([]);
+
 // Non-agent markdown that may live in an agents/ dir.
 const SKIP_FILES = new Set(['README.md', 'MEMORY.md']);
 
@@ -78,17 +140,57 @@ function collectAgentDirs() {
   return dirs;
 }
 
+// A codename is 1-3 words: letters, spaces, apostrophes or hyphens only.
+// Anchored to end-of-line so it can't accidentally swallow a trailing
+// comment or a much longer sentence.
+const CODENAME_CHARS = "[A-Za-z][A-Za-z'.-]*(?:\\s[A-Za-z][A-Za-z'.-]*){0,2}";
+
+// YAML persona field: `name: X` or `name: "X"`, no `- ` list-item prefix
+// (that would be a nested knowledge-base/command entry, not the persona).
+const YAML_NAME_RE = new RegExp(`^[ \\t]*name:[ \\t]*"?(${CODENAME_CHARS})"?[ \\t]*$`, 'm');
+
+// Prose fallback #1: `# Agent: {Name} — {Title}` (common orqx H1 format).
+const PROSE_AGENT_HEADER_RE = new RegExp(`^#\\s*Agent:\\s*(${CODENAME_CHARS})\\s*[—-]`, 'm');
+
+// Prose fallback #2: `# {agent-id} — {Name}` whole-line (e.g. content-orqx.md).
+const PROSE_ID_HEADER_RE = new RegExp(`^#\\s*[\\w./-]+\\s*[—-]\\s*(${CODENAME_CHARS})\\s*$`, 'm');
+
+// Prose fallback #3: markdown identity bullet `- **Nome:** X` / `- **Name:** X`.
+const PROSE_NOME_BULLET_RE = new RegExp(`^-\\s*\\*\\*(?:Nome|Name):\\*\\*\\s*(${CODENAME_CHARS})\\s*$`, 'mi');
+
+/**
+ * Extract the agent codename and where it came from.
+ *
+ * Tries the YAML persona field first (quoted or unquoted); falls back to a
+ * prose declaration for agent files with no YAML block at all (e.g.
+ * design-orqx.md, which declares its codename in a plain markdown list).
+ *
+ * @returns {{ codename: string, source: 'yaml'|'prose' }|null}
+ */
+function extractCodenameWithSource(content) {
+  const yamlMatch = content.match(YAML_NAME_RE);
+  if (yamlMatch) return { codename: yamlMatch[1].trim(), source: 'yaml' };
+
+  const proseMatch =
+    content.match(PROSE_AGENT_HEADER_RE) ||
+    content.match(PROSE_ID_HEADER_RE) ||
+    content.match(PROSE_NOME_BULLET_RE);
+  if (proseMatch) return { codename: proseMatch[1].trim(), source: 'prose' };
+
+  return null;
+}
+
 /**
- * Extract the agent codename: the first quoted `name: "X"` in the file.
+ * Extract just the codename string (back-compat convenience wrapper).
  * @returns {string|null}
  */
 function extractCodename(content) {
-  const match = content.match(/^\s*name:\s*"([^"]+)"\s*$/m);
-  return match ? match[1].trim() : null;
+  const result = extractCodenameWithSource(content);
+  return result ? result.codename : null;
 }
 
 function collectCodenames() {
-  const byName = new Map(); // codename -> [relPath, ...]
+  const byName = new Map(); // codename -> [{ path, source }, ...]
   for (const dir of collectAgentDirs()) {
     let files;
     try {
@@ -104,58 +206,85 @@ function collectCodenames() {
       } catch {
         continue;
       }
-      const codename = extractCodename(content);
-      if (!codename) continue;
+      const found = extractCodenameWithSource(content);
+      if (!found) continue;
       const relPath = path.relative(PROJECT_ROOT, filePath).replace(/\\/g, '/');
-      if (!byName.has(codename)) byName.set(codename, []);
-      byName.get(codename).push(relPath);
+      if (!byName.has(found.codename)) byName.set(found.codename, []);
+      byName.get(found.codename).push({ path: relPath, source: found.source });
     }
   }
   return byName;
 }
 
+/**
+ * @returns {{ blocking: Array, pending: Array }} blocking = real collisions
+ *   this run should fail on; pending = known pre-existing collisions logged
+ *   for visibility but excluded from the exit code (see
+ *   KNOWN_COLLISIONS_PENDING_TRIAGE above).
+ */
 function findCollisions(byName) {
-  const collisions = [];
-  for (const [codename, files] of byName) {
-    if (files.length > 1 && !ALLOWED_DUPLICATES.has(codename)) {
-      collisions.push({ codename, files: files.sort() });
+  const blocking = [];
+  const pending = [];
+  for (const [codename, entries] of byName) {
+    if (entries.length <= 1) continue;
+    const files = [...entries].sort((a, b) => a.path.localeCompare(b.path));
+    if (ALLOWED_DUPLICATES.has(codename)) continue;
+    if (KNOWN_COLLISIONS_PENDING_TRIAGE.has(codename)) {
+      pending.push({ codename, files });
+      continue;
     }
+    blocking.push({ codename, files });
   }
-  return collisions.sort((a, b) => a.codename.localeCompare(b.codename));
+  const byCodename = (a, b) => a.codename.localeCompare(b.codename);
+  return { blocking: blocking.sort(byCodename), pending: pending.sort(byCodename) };
 }
 
-function formatReport(byName, collisions) {
+function formatReport(byName, { blocking, pending }) {
   const lines = [];
-  const totalAgents = [...byName.values()].reduce((n, f) => n + f.length, 0);
+  const totalAgents = [...byName.values()].reduce((n, entries) => n + entries.length, 0);
   lines.push(`Scanned ${totalAgents} agent file(s) with a codename; ${byName.size} distinct codename(s).`);
-  if (collisions.length === 0) {
-    lines.push('OK — every codename is unique (master dual-register allow-listed).');
-    return lines.join('\n');
+
+  if (blocking.length === 0) {
+    lines.push('OK — every codename is unique (master dual-register + known pending debt excluded).');
+  } else {
+    lines.push('');
+    lines.push(`FAIL — ${blocking.length} duplicate codename(s) found:`);
+    for (const c of blocking) {
+      lines.push(`  "${c.codename}" used by ${c.files.length} agents:`);
+      for (const f of c.files) lines.push(`    - ${f.path}${f.source === 'prose' ? ' (prose)' : ''}`);
+    }
+    lines.push('');
+    lines.push('Fix: rename the least-canonical side to a unique codename.');
   }
-  lines.push('');
-  lines.push(`FAIL — ${collisions.length} duplicate codename(s) found:`);
-  for (const c of collisions) {
-    lines.push(`  "${c.codename}" used by ${c.files.length} agents:`);
-    for (const f of c.files) lines.push(`    - ${f}`);
+
+  if (pending.length > 0) {
+    lines.push('');
+    lines.push(
+      `NOTE — ${pending.length} known pre-existing collision(s) NOT blocking this run ` +
+        '(tracked debt, see KNOWN_COLLISIONS_PENDING_TRIAGE — schema uniformization backlog):',
+    );
+    for (const c of pending) {
+      lines.push(`  "${c.codename}" used by ${c.files.length} agents:`);
+      for (const f of c.files) lines.push(`    - ${f.path}${f.source === 'prose' ? ' (prose)' : ''}`);
+    }
   }
-  lines.push('');
-  lines.push('Fix: rename the least-canonical side to a unique codename.');
+
   return lines.join('\n');
 }
 
 function main() {
   const args = parseArgs();
   const byName = collectCodenames();
-  const collisions = findCollisions(byName);
+  const { blocking, pending } = findCollisions(byName);
 
   if (args.json) {
-    console.log(JSON.stringify({ collisions, distinct: byName.size }, null, 2));
+    console.log(JSON.stringify({ collisions: blocking, pendingCollisions: pending, distinct: byName.size }, null, 2));
   } else if (!args.quiet) {
     console.log('=== validate-agent-codenames ===');
-    console.log(formatReport(byName, collisions));
+    console.log(formatReport(byName, { blocking, pending }));
   }
 
-  return collisions.length > 0 ? 1 : 0;
+  return blocking.length > 0 ? 1 : 0;
 }
 
 if (require.main === module) {
@@ -165,10 +294,12 @@ if (require.main === module) {
 module.exports = {
   parseArgs,
   extractCodename,
+  extractCodenameWithSource,
   collectAgentDirs,
   collectCodenames,
   findCollisions,
   formatReport,
   main,
   ALLOWED_DUPLICATES,
+  KNOWN_COLLISIONS_PENDING_TRIAGE,
 };

package/scripts/validate-agents-md.js

@@ -27,6 +27,7 @@
 
 const fs = require('fs');
 const path = require('path');
+const { atomicWriteSync } = require('../.sinapse-ai/core/synapse/utils/atomic-write');
 const { execSync } = require('child_process');
 
 const ROOT = path.resolve(__dirname, '..');
@@ -113,7 +114,7 @@ function main() {
   }
 
   if (FIX) {
-    fs.writeFileSync(AGENTS_MD, fixed, 'utf8');
+    atomicWriteSync(AGENTS_MD, fixed, 'utf8');
     console.log('AGENTS.md count(s) rewritten to match disk:');
     for (const mm of mismatches) console.log(`  - ${mm}`);
     process.exit(0);

package/scripts/validate-article-xi.js

@@ -31,7 +31,7 @@
 'use strict';
 
 const path = require('path');
-const { execSync } = require('child_process');
+const { execFileSync } = require('child_process');
 
 const ROOT = path.resolve(__dirname, '..');
 
@@ -61,9 +61,11 @@ function getBaseRef(args) {
   return 'main';
 }
 
-function gitExec(cmd) {
+// SECURITY: args go straight to git without a shell — baseRef comes from CLI
+// or GITHUB_BASE_REF and must never be interpolated into a shell string.
+function gitExec(args) {
   try {
-    return execSync(cmd, { cwd: ROOT, stdio: ['ignore', 'pipe', 'pipe'] }).toString();
+    return execFileSync('git', args, { cwd: ROOT, stdio: ['ignore', 'pipe', 'pipe'] }).toString();
   } catch (_err) {
     return null;
   }
@@ -78,14 +80,14 @@ function gitExec(cmd) {
  */
 function diffAgainstBase(baseRef) {
   // Try fetching first (CI may need it). Silent if it fails.
-  gitExec(`git fetch origin ${baseRef} --quiet`);
+  gitExec(['fetch', 'origin', baseRef, '--quiet']);
 
   // Use raw diff with --name-status -M (rename detection).
   const tripleDot = `origin/${baseRef}...HEAD`;
-  let raw = gitExec(`git diff --name-status -M ${tripleDot}`);
+  let raw = gitExec(['diff', '--name-status', '-M', tripleDot]);
   if (raw === null) {
     // Fallback: local base ref
-    raw = gitExec(`git diff --name-status -M ${baseRef}...HEAD`);
+    raw = gitExec(['diff', '--name-status', '-M', `${baseRef}...HEAD`]);
   }
   if (raw === null) {
     return { error: `nao foi possivel comparar com ${baseRef}`, deletions: [], renames: [] };
@@ -110,8 +112,8 @@ function diffAgainstBase(baseRef) {
 
 function getOverrideText(baseRef) {
   // Aggregate commit messages between base..HEAD + PR body.
-  const commitMsgs = gitExec(`git log --format=%B origin/${baseRef}..HEAD`)
-    || gitExec(`git log --format=%B ${baseRef}..HEAD`)
+  const commitMsgs = gitExec(['log', '--format=%B', `origin/${baseRef}..HEAD`])
+    || gitExec(['log', '--format=%B', `${baseRef}..HEAD`])
     || '';
   const prBody = process.env.GITHUB_PR_BODY || '';
   const forcedOverride = process.env.ARTICLE_XI_OVERRIDE || '';

package/scripts/validate-install-docs.js

@@ -13,9 +13,11 @@
  *   npx sinapse-ai install
  *   npx sinapse-ai update
  *   npx sinapse-ai uninstall
+ *   npx sinapse-ai agents
+ *   npx sinapse-ai ideate
  *
  * Non-canonical patterns the guard rejects in public docs:
- *   - `sinapse install|update|uninstall`  (missing the `-ai` suffix)
+ *   - `sinapse install|update|uninstall|agents|ideate`  (missing the `-ai` suffix)
  *   - `sinapse-minimal`                    (internal/legacy binary)
  *   - `sinapse-graph`                      (internal/legacy binary)
  *   - `install-squads.sh`                  (internal script, must not be exposed)
@@ -69,6 +71,19 @@ const FORBIDDEN_PATTERNS = [
     regex: /\bsinapse\s+uninstall\b/,
     reason: 'Use `npx sinapse-ai uninstall` (the `sinapse uninstall` form is deprecated).',
   },
+  {
+    // Onda1-S1 (audit AF-20260702 item 1.2) — `agents`/`ideate` are owned by
+    // bin/cli.js (the `sinapse-ai` binary); same canonical-prefix rule as
+    // install/update/uninstall above.
+    id: 'legacy-sinapse-agents',
+    regex: /\bsinapse\s+agents\b/,
+    reason: 'Use `npx sinapse-ai agents` (the `sinapse agents` form is not the canonical public entry point).',
+  },
+  {
+    id: 'legacy-sinapse-ideate',
+    regex: /\bsinapse\s+ideate\b/,
+    reason: 'Use `npx sinapse-ai ideate` (the `sinapse ideate` form is not the canonical public entry point).',
+  },
   {
     id: 'sinapse-minimal-binary',
     regex: /\bsinapse-minimal\b/,
@@ -113,8 +128,12 @@ const PUBLIC_DOC_DIRS = ['docs/installation'];
 
 /**
  * Additional top-level public doc files (relative to root).
+ *
+ * Onda1-S1 (audit AF-20260702 item 1.2) — the EN/PT user guides teach the
+ * Quick Start commands (including `agents`/`ideate`) and are exactly where
+ * the drift was found; bring them under the same guard as README.
  */
-const PUBLIC_DOC_FILES = ['README.md'];
+const PUBLIC_DOC_FILES = ['README.md', 'docs/guides/user-guide.md', 'docs/pt/guides/user-guide.md'];
 
 /**
  * Path fragments that identify historical / release notes. Any file whose
@@ -317,6 +336,8 @@ function formatReport(result) {
   lines.push('  npx sinapse-ai install');
   lines.push('  npx sinapse-ai update');
   lines.push('  npx sinapse-ai uninstall');
+  lines.push('  npx sinapse-ai agents');
+  lines.push('  npx sinapse-ai ideate');
   lines.push('');
   lines.push('Wrap developer-only mentions in an `### Internal (developer-only)` section to opt them out.');
   lines.push('');

package/scripts/validate-no-external-refs.js

@@ -240,21 +240,27 @@ function isBinaryFile(absPath) {
 function scanFile(rootDir, relPath) {
   if (isAllowListed(relPath)) return [];
   const absPath = path.join(rootDir, relPath);
-  let stat;
+
+  // Size cap and read share one fd, so the scanned bytes are the sized bytes.
+  let content;
+  let fd;
   try {
-    stat = fs.statSync(absPath);
+    fd = fs.openSync(absPath, 'r');
   } catch {
     return [];
   }
-  if (!stat.isFile()) return [];
-  if (stat.size > MAX_SCAN_BYTES) return [];
-  if (isBinaryFile(absPath)) return [];
-
-  let content;
   try {
-    content = fs.readFileSync(absPath, 'utf8');
+    const stat = fs.fstatSync(fd);
+    if (!stat.isFile()) return [];
+    if (stat.size > MAX_SCAN_BYTES) return [];
+    if (isBinaryFile(absPath)) return [];
+    const buf = Buffer.alloc(stat.size);
+    fs.readSync(fd, buf, 0, stat.size, 0);
+    content = buf.toString('utf8');
   } catch {
     return [];
+  } finally {
+    fs.closeSync(fd);
   }
 
   const lines = content.split(/\r?\n/);

package/sinapse/agents/sinapse-orqx.md

@@ -360,7 +360,7 @@ routing_table:
 
     - squad: claude-code-mastery
       prefix: claude
-      orchestrator: swarm-orqx (Nexus)
+      orchestrator: swarm-orqx (Relay)
       invocation: "/claude:agents:swarm-orqx"
       domain: "Claude Code mastery, prompt engineering, MCP, automacao, hooks, skills, plugins, agent teams, context engineering"
       agents: 8
@@ -652,7 +652,7 @@ relationships:
       context: "All product strategy, discovery, roadmap"
     - agent: research-orqx (Prism)
       context: "All market research, competitive intelligence"
-    - agent: swarm-orqx (Nexus)
+    - agent: swarm-orqx (Relay)
       context: "All Claude Code mastery, prompt engineering, MCP"
     - agent: council-orqx (Zenith)
       context: "All strategic counsel, mental models, advisory"
@@ -766,7 +766,7 @@ Imperator can provide ecosystem-wide insights by combining capabilities across s
 | 9 | paidmedia | pm | Apex | Midia paga, Meta/Google Ads |
 | 10 | product | product | Vector | Produto, discovery, roadmap |
 | 11 | research | research | Prism | Pesquisa, inteligencia competitiva |
-| 12 | claude-code-mastery | claude | Nexus | Claude Code, prompt engineering |
+| 12 | claude-code-mastery | claude | Relay | Claude Code, prompt engineering |
 | 13 | council | council | Zenith | Conselho estrategico, advisory |
 | 14 | storytelling | narrative | Arc | Storytelling, pitch, apresentacao |
 | 15 | cybersecurity | cyber | Fortress | Cybersecurity, compliance, pentest |

package/sinapse/agents/snps-orqx.md

@@ -360,7 +360,7 @@ routing_table:
 
     - squad: claude-code-mastery
       prefix: claude
-      orchestrator: swarm-orqx (Nexus)
+      orchestrator: swarm-orqx (Relay)
       invocation: "/claude:agents:swarm-orqx"
       domain: "Claude Code mastery, prompt engineering, MCP, automacao, hooks, skills, plugins, agent teams, context engineering"
       agents: 8
@@ -652,7 +652,7 @@ relationships:
       context: "All product strategy, discovery, roadmap"
     - agent: research-orqx (Prism)
       context: "All market research, competitive intelligence"
-    - agent: swarm-orqx (Nexus)
+    - agent: swarm-orqx (Relay)
       context: "All Claude Code mastery, prompt engineering, MCP"
     - agent: council-orqx (Zenith)
       context: "All strategic counsel, mental models, advisory"
@@ -766,7 +766,7 @@ Imperator can provide ecosystem-wide insights by combining capabilities across s
 | 9 | paidmedia | pm | Apex | Midia paga, Meta/Google Ads |
 | 10 | product | product | Vector | Produto, discovery, roadmap |
 | 11 | research | research | Prism | Pesquisa, inteligencia competitiva |
-| 12 | claude-code-mastery | claude | Nexus | Claude Code, prompt engineering |
+| 12 | claude-code-mastery | claude | Relay | Claude Code, prompt engineering |
 | 13 | council | council | Zenith | Conselho estrategico, advisory |
 | 14 | storytelling | narrative | Arc | Storytelling, pitch, apresentacao |
 | 15 | cybersecurity | cyber | Fortress | Cybersecurity, compliance, pentest |

package/sinapse/knowledge-base/routing-catalog.md

@@ -17,7 +17,7 @@
 | 9 | squad-paidmedia | pm | paidmedia-orqx (Apex) | `/pm:agents:paidmedia-orqx` |
 | 10 | squad-product | product | product-orqx (Vector) | `/product:agents:product-orqx` |
 | 11 | squad-research | research | research-orqx (Prism) | `/research:agents:research-orqx` |
-| 12 | claude-code-mastery | claude | swarm-orqx (Nexus) | `/claude:agents:swarm-orqx` |
+| 12 | claude-code-mastery | claude | swarm-orqx (Relay) | `/claude:agents:swarm-orqx` |
 | 13 | squad-council | council | council-orqx (Zenith) | `/council:agents:council-orqx` |
 | 14 | squad-storytelling | narrative | storytelling-orqx (Arc) | `/narrative:agents:storytelling-orqx` |
 | 15 | squad-cybersecurity | cyber | cyber-orqx (Fortress) | `/cyber:agents:cyber-orqx` |

package/sinapse/tasks/diagnose-and-route.md

@@ -63,7 +63,7 @@ Mapear contra a routing table:
 | Midia paga/ads | paid-media | pm | Apex | `/pm:agents:paidmedia-orqx` |
 | Produto/discovery | product-systems | product | Vector | `/product:agents:product-orqx` |
 | Pesquisa/competitivo | research-intelligence | research | Prism | `/research:agents:research-orqx` |
-| Claude Code/prompts | claude-mastery | claude | Nexus | `/claude:agents:swarm-orqx` |
+| Claude Code/prompts | claude-mastery | claude | Relay | `/claude:agents:swarm-orqx` |
 | Estrategia/advisory | strategic-council | council | Zenith | `/council:agents:council-orqx` |
 | Narrativa/pitch | narrative-masters | narrative | Arc | `/narrative:agents:storytelling-orqx` |
 | Seguranca/compliance | cyber-defense | cyber | Fortress | `/cyber:agents:cyber-orqx` |

package/squads/claude-code-mastery/CHANGELOG.md

@@ -5,7 +5,7 @@ All notable changes to the Claude Code Mastery squad.
 ## [1.0.0] - 2026-03-02
 
 ### Added
-- 8 specialist agents: claude-mastery-chief (Orion), hooks-architect (Latch), mcp-integrator (Piper), swarm-orqx (Nexus), config-engineer (Sigil), skill-craftsman (Anvil), project-integrator (Conduit), roadmap-sentinel (Vigil)
+- 8 specialist agents: claude-mastery-chief (Orion), hooks-architect (Latch), mcp-integrator (Piper), swarm-orqx (Relay), config-engineer (Sigil), skill-craftsman (Anvil), project-integrator (Conduit), roadmap-sentinel (Vigil)
 - 26 executable tasks across all agents
 - 3 multi-phase workflows (wf-project-setup, wf-knowledge-update, wf-audit-complete)
 - 5 knowledge base files (quick-ref, project-type-signatures, hook-patterns, ci-cd-patterns, mcp-catalog)

package/squads/claude-code-mastery/README.md

@@ -13,7 +13,7 @@
             ┌─────────┬───────┴───────┬──────────┐
             |         |               |          |
     ┌───────┴──┐  ┌───┴────┐  ┌──────┴───┐  ┌───┴──────┐
-    |  Latch   |  | Piper  |  |  Nexus   |  |  Sigil   |
+    |  Latch   |  | Piper  |  |  Relay   |  |  Sigil   |
     |  Hooks   |  |  MCP   |  |  Swarm   |  |  Config  |
     | Tier 1   |  | Tier 1 |  |  Tier 1  |  |  Tier 1  |
     └──────────┘  └────────┘  └──────────┘  └──────────┘
@@ -32,7 +32,7 @@
 | 0 | claude-mastery-chief | Orion | Original | 554 | Triage, routing, cross-cutting knowledge |
 | 1 | hooks-architect | Latch | disler (IndyDevDan) | 1,013 | 17 hook events, automation, damage control |
 | 1 | mcp-integrator | Piper | Peter Steinberger (@steipete) | 791 | MCP servers, tool discovery, context budget |
-| 1 | swarm-orqx | Nexus | Kieran Klaassen + Reuven Cohen | 1,008 | Agent teams, subagents, parallel execution |
+| 1 | swarm-orqx | Relay | Kieran Klaassen + Reuven Cohen | 1,008 | Agent teams, subagents, parallel execution |
 | 1 | config-engineer | Sigil | SuperClaude-Org | 663 | Settings, permissions, CLAUDE.md, sandbox |
 | 2 | skill-craftsman | Anvil | spec-driven agile-AI methodology | 1,046 | Skills, plugins, commands, context engineering |
 | 2 | project-integrator | Conduit | Daniel Miessler (PAI) | 959 | Project integration, CI/CD, SINAPSE bridge |
@@ -66,7 +66,7 @@ Or use the SINAPSE activation:
 |-------------------|-----------|-------------|
 | Hooks (17 events) | Latch | `*create-hook`, `*audit-hooks`, `*hook-patterns` |
 | MCP Integration | Piper | `*add-server`, `*audit-mcp`, `*create-mcp-server` |
-| Subagents & Teams | Nexus | `*create-agent`, `*create-team`, `*orchestrate` |
+| Subagents & Teams | Relay | `*create-agent`, `*create-team`, `*orchestrate` |
 | Settings & Permissions | Sigil | `*configure`, `*permission-strategy`, `*sandbox-setup` |
 | Skills & Plugins | Anvil | `*create-skill`, `*create-plugin`, `*context-strategy` |
 | Project Integration | Conduit | `*integrate-project`, `*brownfield-setup`, `*ci-cd-setup` |
@@ -94,9 +94,9 @@ This squad understands both Claude Code native capabilities AND the SINAPSE fram
 
 | SINAPSE Concept | Claude Code Equivalent | Bridge Agent |
 |-------------|----------------------|-------------|
-| Agents (@developer, @qa...) | Subagents (.claude/agents/) | Nexus |
+| Agents (@developer, @qa...) | Subagents (.claude/agents/) | Relay |
 | Tasks (.sinapse-ai/tasks/) | Skills (.claude/skills/) | Anvil |
-| Workflows | Multi-step sessions | Nexus + Orion |
+| Workflows | Multi-step sessions | Relay + Orion |
 | core-config.yaml | .claude/settings.json | Sigil |
 | Python hooks (monitor/) | Native hooks (command/http/prompt/agent) | Latch |
 | Quality gates | Hook-based validation | Latch + Sigil |
@@ -112,7 +112,7 @@ squads/claude-code-mastery/
 │   ├── claude-mastery-chief.md    # Tier 0: Orchestrator (Orion)
 │   ├── hooks-architect.md         # Tier 1: Hooks (Latch)
 │   ├── mcp-integrator.md          # Tier 1: MCP (Piper)
-│   ├── swarm-orqx.md      # Tier 1: Subagents/Teams (Nexus)
+│   ├── swarm-orqx.md      # Tier 1: Subagents/Teams (Relay)
 │   ├── config-engineer.md         # Tier 1: Settings/Config (Sigil)
 │   ├── skill-craftsman.md         # Tier 2: Skills/Plugins (Anvil)
 │   ├── project-integrator.md      # Tier 2: Integration (Conduit)