pan-wizard 3.31.0 → 3.33.0

package/pan-wizard-core/bin/lib/config.cjs

@@ -103,6 +103,102 @@ function buildConfigDefaults(hasBraveSearch, userDefaults) {
   };
 }
 
+// What PAN reads beyond the keys buildConfigDefaults writes (market-ideas M26). The
+// rest of the known set is derived from the defaults, so a new default is known the
+// day it is added. Each entry has a reader: loadConfig's flat aliases and alias
+// sections (`planning.*`, `git.*`), the stop guard (`mode`, `workflow.stop_guard`),
+// the roadmapper and new-project (`depth`), memory-optimize (`memory`), and the
+// workflows that config-get `workflow.auto_advance`. tests/config-keys in
+// tests/config.test.cjs pins this list against the keys shipped code and prose use.
+const CONFIG_KEYS_BEYOND_DEFAULTS = Object.freeze({
+  top: Object.freeze(['research', 'plan_checker', 'verifier', 'nyquist_validation', 'mode', 'depth', 'planning', 'git',
+    'focus', 'model_overrides', 'effort_overrides', 'cost', 'cache', 'build', 'verification', 'concurrency', 'memory']),
+  sections: Object.freeze({
+    workflow: Object.freeze(['auto_advance', 'stop_guard']),
+    planning: Object.freeze(['commit_docs', 'search_gitignored']),
+    git: Object.freeze(['branching_strategy', 'phase_branch_template', 'milestone_branch_template']),
+  }),
+});
+// Sections whose keys are the user's own (model ids, agent names, file lists,
+// commands), so nothing below them is checked.
+const OPEN_CONFIG_SECTIONS = Object.freeze(['focus', 'model_overrides', 'effort_overrides', 'cost', 'cache', 'build', 'verification', 'concurrency', 'memory']);
+
+/** The known top-level keys, and the known keys one level inside each closed section. */
+function knownConfigKeys() {
+  const defaults = buildConfigDefaults(false, {});
+  const top = new Set([...Object.keys(defaults), ...CONFIG_KEYS_BEYOND_DEFAULTS.top]);
+  const sections = new Map();
+  for (const [k, v] of Object.entries(defaults)) {
+    if (v && typeof v === 'object' && !Array.isArray(v)) sections.set(k, new Set(Object.keys(v)));
+  }
+  for (const [k, list] of Object.entries(CONFIG_KEYS_BEYOND_DEFAULTS.sections)) {
+    sections.set(k, new Set([...(sections.get(k) || []), ...list]));
+  }
+  for (const k of OPEN_CONFIG_SECTIONS) sections.delete(k);
+  return { top, sections };
+}
+
+/**
+ * Whether PAN reads a dot-notation key path: the top-level key is known, and inside a
+ * closed section so is the second segment. Deeper segments are not checked.
+ * @param {string} keyPath
+ * @returns {boolean}
+ */
+function isKnownConfigPath(keyPath) {
+  const [head, sub] = String(keyPath).split('.');
+  const { top, sections } = knownConfigKeys();
+  if (!top.has(head)) return false;
+  return sub === undefined || !sections.has(head) || sections.get(head).has(sub);
+}
+
+/**
+ * The key paths in a parsed config.json that PAN does not read (market-ideas M26):
+ * unknown top-level keys, and unknown keys one level inside a closed section.
+ * @param {Object} config
+ * @returns {string[]}
+ */
+function unknownConfigKeys(config) {
+  if (!config || typeof config !== 'object' || Array.isArray(config)) return [];
+  const { top, sections } = knownConfigKeys();
+  const unknown = [];
+  for (const [k, v] of Object.entries(config)) {
+    if (!top.has(k)) { unknown.push(k); continue; }
+    const known = sections.get(k);
+    if (!known || !v || typeof v !== 'object' || Array.isArray(v)) continue;
+    for (const sub of Object.keys(v)) if (!known.has(sub)) unknown.push(`${k}.${sub}`);
+  }
+  return unknown;
+}
+
+/** The known key path nearest an unknown one (edit distance 2 or less), or null. */
+function nearestConfigKey(keyPath) {
+  const { top, sections } = knownConfigKeys();
+  const candidates = [...top];
+  for (const [k, subs] of sections) for (const s of subs) candidates.push(`${k}.${s}`);
+  const want = String(keyPath).split('.').slice(0, 2).join('.');
+  let best = null;
+  let bestD = 3;
+  for (const c of candidates) {
+    const d = editDistance(want, c);
+    if (d < bestD) { best = c; bestD = d; }
+  }
+  return best;
+}
+
+function editDistance(a, b) {
+  const row = Array.from({ length: b.length + 1 }, (_, j) => j);
+  for (let i = 1; i <= a.length; i++) {
+    let prev = row[0];
+    row[0] = i;
+    for (let j = 1; j <= b.length; j++) {
+      const tmp = row[j];
+      row[j] = Math.min(row[j] + 1, row[j - 1] + 1, prev + (a[i - 1] === b[j - 1] ? 0 : 1));
+      prev = tmp;
+    }
+  }
+  return row[b.length];
+}
+
 function cmdConfigEnsureSection(cwd, raw) {
   const configPath = path.join(planningPath(cwd), CONFIG_FILE);
 
@@ -216,6 +312,14 @@ function cmdConfigSet(cwd, keyPath, value, raw) {
   try {
     fs.writeFileSync(configPath, JSON.stringify(config, null, 2), 'utf-8');
     const result = { updated: true, key: keyPath, value: parsedValue };
+    // A key PAN does not read is still written (a newer PAN, or the user's own
+    // note, may want it), but never silently: a typo used to look like a setting
+    // that took effect (market-ideas M26).
+    if (!isKnownConfigPath(keyPath)) {
+      const near = nearestConfigKey(keyPath);
+      result.warning = `unknown config key "${keyPath}": PAN does not read it` + (near ? `; did you mean "${near}"?` : '');
+      process.stderr.write(`pan-tools: ${result.warning}\n`);
+    }
     output(result, raw, `${keyPath}=${parsedValue}`);
   } catch (err) {
     // Config file could not be written — disk full, permissions, etc.
@@ -664,6 +768,12 @@ function cmdStandardsTools(cwd, standardId, raw) {
 
 module.exports = {
   buildConfigDefaults,
+  CONFIG_KEYS_BEYOND_DEFAULTS,
+  OPEN_CONFIG_SECTIONS,
+  knownConfigKeys,
+  isKnownConfigPath,
+  unknownConfigKeys,
+  nearestConfigKey,
   cmdConfigEnsureSection,
   cmdConfigSet,
   cmdConfigGet,

package/pan-wizard-core/bin/lib/constants.cjs

@@ -207,16 +207,18 @@ const STANDARDS_CATALOG = {
     level: 'foundational',
     url: 'https://owasp.org/www-project-top-ten/',
     checklist: [
-      'A01: Broken Access Control — verify authorization checks on all endpoints',
-      'A02: Cryptographic Failures — verify sensitive data encryption at rest and in transit',
-      'A03: Injection — verify input validation and parameterized queries',
-      'A04: Insecure Design — verify threat modeling and secure design patterns',
-      'A05: Security Misconfiguration — verify default credentials removed, headers set',
-      'A06: Vulnerable Components — verify dependency scanning, no known CVEs',
+      // The 2025 list (top10.owasp.org): until 2026-10-03 this held the 2021 list
+      // under the 2025 name (prompt audit, market-ideas M34). SSRF is now part of A01.
+      'A01: Broken Access Control — verify authorization checks on all endpoints, and allowlists for server-side requests (SSRF)',
+      'A02: Security Misconfiguration — verify default credentials removed, headers set',
+      'A03: Software Supply Chain Failures — verify dependency scanning, no known CVEs, pinned and verified packages and build tooling',
+      'A04: Cryptographic Failures — verify sensitive data encryption at rest and in transit',
+      'A05: Injection — verify input validation and parameterized queries',
+      'A06: Insecure Design — verify threat modeling and secure design patterns',
       'A07: Authentication Failures — verify MFA support, session management, credential storage',
-      'A08: Software and Data Integrity — verify CI/CD pipeline integrity, signed updates',
-      'A09: Logging and Monitoring — verify security events logged, alerts configured',
-      'A10: SSRF — verify server-side request validation, allowlists',
+      'A08: Software or Data Integrity Failures — verify CI/CD pipeline integrity, signed updates',
+      'A09: Security Logging and Alerting Failures — verify security events logged, alerts configured',
+      'A10: Mishandling of Exceptional Conditions — verify errors fail closed, are handled, and release resources',
     ],
   },
   'owasp-asvs-l1': {
@@ -541,6 +543,15 @@ const HEALTH_STATUS = { HEALTHY: 'healthy', DEGRADED: 'degraded', BROKEN: 'broke
 /** Max JSON payload size before writing to tmpfile (bytes) */
 const MAX_JSON_SIZE = 50000;
 
+/**
+ * Shape version of the machine-readable state JSON: `state` (load), `state json`,
+ * `progress` (json) and the MCP resources that wrap them (market item M17).
+ * Additive rule: fields may be added within a major version and are never renamed
+ * or removed; readers accept any 1.x and ignore fields they do not know. A rename
+ * or removal is a major bump ("2.0").
+ */
+const STATE_CONTRACT = '1.0';
+
 /** Width of progress bar in status display (character count) */
 const PROGRESS_BAR_WIDTH = 10;
 
@@ -724,6 +735,24 @@ const FOREIGN_PLANNING_MARKERS = Object.freeze({
       'graphify.enabled', 'intel.enabled', 'hooks.workflow_guard', 'dynamic_routing', 'resolve_model_ids',
     ]),
   }),
+  // planning-with-files (OthmanAdi/planning-with-files README, read 2026-09-28 at
+  // v3.21.0; market item M3 / MI-028). It writes INTO .planning/ beside whatever else
+  // lives there: `.active_plan`, `.attestation`, `sessions/`, `ledger-<agent>.jsonl`, and
+  // one `YYYY-MM-DD-slug/` directory per task holding task_plan.md / findings.md /
+  // progress.md. PAN writes none of those names. `coexists`: unlike gsd-core, a match
+  // means the tree is SHARED, not owned. PAN keeps working in it and says so, rather
+  // than refusing. planning-with-files gitignores .planning/ by default, which is what
+  // the I005 health check exists for.
+  planning_with_files: Object.freeze({
+    tool: 'planning-with-files',
+    coexists: true,
+    files: Object.freeze(['.active_plan', '.attestation']),
+    filePatterns: Object.freeze([/^ledger-[A-Za-z0-9._-]+\.jsonl$/]),
+    dirs: Object.freeze(['sessions']),
+    // A dated task directory is strong evidence only with the tool's plan file in it.
+    dirPatterns: Object.freeze([{ re: /^\d{4}-\d{2}-\d{2}-[A-Za-z0-9][A-Za-z0-9._-]*$/, requires: 'task_plan.md' }]),
+    configKeys: Object.freeze([]),
+  }),
 });
 
 module.exports = {
@@ -806,6 +835,7 @@ module.exports = {
   HEALTH_STATUS,
   FOREIGN_PLANNING_MARKERS,
   MAX_JSON_SIZE,
+  STATE_CONTRACT,
   PROGRESS_BAR_WIDTH,
   MAX_SLUG_LENGTH,
   FILLED_BLOCK,

package/pan-wizard-core/bin/lib/core.cjs

@@ -39,8 +39,13 @@ const { planningPath, planningRel } = require('./utils.cjs');
 // budget-profile project on Codex or OpenCode was told to spawn a model called
 // "mid" (reality check R28).
 //   - openai: Codex's subagent docs (learn.chatgpt.com/docs/agent-configuration/
-//     subagents, read 2026-09-23): "start with gpt-6-sol. Use gpt-6-luna when you
-//     want a faster, lower-cost option". gpt-6-astra is the flagship above both.
+//     subagents, re-read 2026-10-03) now start demanding agents on gpt-6.1-sol "when
+//     your signed-in account or workspace has access", and keep gpt-6-luna as the
+//     "faster, lower-cost option". The mid tier stays gpt-6-sol: GPT-6.1 Sol's launch
+//     keeps it off on Enterprise and Edu until an administrator enables it
+//     (learn.chatgpt.com/docs/models, read 2026-10-03), so pinning it would hand
+//     those users a model they cannot run. The inherit tier runs Codex's own
+//     default, which is gpt-6.1-sol where available. gpt-6-astra is the flagship.
 //   - google: the newest stable Flash and Flash-Lite on ai.google.dev's models page
 //     (read 2026-09-23). The 2.5 family these rows used is "not deprecated" but
 //     limited to users who have used it before, so a new project may not reach it.
@@ -58,12 +63,14 @@ const PROVIDER_MODELS = {
 // packages/web/src/content/docs/models.mdx and models.dev/api.json, both read
 // 2026-09-26; every id below exists there under that provider). Same tiers as
 // PROVIDER_MODELS, provider-qualified; `default` is Anthropic as it is there.
-// Reality check R41.
+// Reality check R41. The Anthropic mid tier follows the `sonnet` alias, which
+// resolves to Sonnet 5.5 on the Anthropic API from Claude Code 2.1.284;
+// models.dev lists `anthropic/claude-sonnet-5-5` (api.json, read 2026-10-03).
 const OPENCODE_MODELS = {
-  anthropic: { reasoning: 'inherit', mid: 'anthropic/claude-sonnet-5', fast: 'anthropic/claude-haiku-4-5' },
-  openai:    { reasoning: 'inherit', mid: 'openai/gpt-6-sol',          fast: 'openai/gpt-6-luna' },
-  google:    { reasoning: 'inherit', mid: 'google/gemini-3.8-flash',   fast: 'google/gemini-3.5-flash-lite' },
-  default:   { reasoning: 'inherit', mid: 'anthropic/claude-sonnet-5', fast: 'anthropic/claude-haiku-4-5' },
+  anthropic: { reasoning: 'inherit', mid: 'anthropic/claude-sonnet-5-5', fast: 'anthropic/claude-haiku-4-5' },
+  openai:    { reasoning: 'inherit', mid: 'openai/gpt-6-sol',            fast: 'openai/gpt-6-luna' },
+  google:    { reasoning: 'inherit', mid: 'google/gemini-3.8-flash',     fast: 'google/gemini-3.5-flash-lite' },
+  default:   { reasoning: 'inherit', mid: 'anthropic/claude-sonnet-5-5', fast: 'anthropic/claude-haiku-4-5' },
 };
 
 /**
@@ -552,6 +559,18 @@ function escapeRegex(value) {
   return String(value).replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
 }
 
+/**
+ * Make free text safe for one markdown table cell: escape backslashes and pipes,
+ * and fold every run of line breaks into one space. A raw `|` splits the row into
+ * an extra column, and state.md's tables are appended to for the life of a
+ * project, so one bad row breaks every later read of the table (market-ideas M21).
+ * @param {*} value - Cell text; null and undefined become ''
+ * @returns {string}
+ */
+function escapeTableCell(value) {
+  return String(value ?? '').replace(/\\/g, '\\\\').replace(/\|/g, '\\|').replace(/[\r\n]+/g, ' ');
+}
+
 /**
  * Normalize a phase identifier to zero-padded format.
  * Examples: "1" → "01", "3A" → "03A", "12.1" → "12.1"
@@ -1454,6 +1473,7 @@ module.exports = {
   isGitRepo,
   execGit,
   escapeRegex,
+  escapeTableCell,
   normalizePhaseName,
   comparePhaseNum,
   searchPhaseInDir,

package/pan-wizard-core/bin/lib/cost.cjs

@@ -59,7 +59,7 @@ const DEFAULT_RATES = {
   // Anthropic — platform.claude.com/docs/en/about-claude/pricing, every row read
   // 2026-09-23. Opus 4.6+/Opus 5 are $5/$25 (the old $15/$75 Opus pricing ended with
   // the 4.5 generation). cache_write is the 5-minute write rate (1.25× input; the
-  // 1-hour write bills 2×, which the ledger does not split yet). Cache reads follow
+  // 1-hour write bills 2× and is priced from cache_write_1h, see above). Cache reads follow
   // the 0.1× convention EXCEPT on three models the page footnotes: Fable 5.1 and
   // Mythos 5.1 at 0.025× input, Opus 5.5 at 0.05×. Each needs its own row, because
   // the family-prefix fallback would otherwise price it at its predecessor's reads.
@@ -91,6 +91,12 @@ const DEFAULT_RATES = {
   // 2026-08-31 was made permanent and the scheduled rise to $3/$15 cancelled.
   // Lesson: never write down a pre-announced price — this row carried the future
   // rate for a month and over-billed by half.
+  // Sonnet 5.5 (claude-sonnet-5-5): the same prices as Sonnet 5 — $2/$10, 5-minute
+  // writes $2.50, 1-hour $4, hits $0.20 (pricing page, read 2026-10-03). Claude Code
+  // 2.1.284 points the `sonnet` alias at it on the Anthropic API (model-config, read
+  // 2026-10-03). The family prefix already priced it at Sonnet 5's identical row; its
+  // own row means a later price change to either model cannot misprice the other.
+  'claude-sonnet-5-5':  { input: 2.0,  output: 10.0, cache_read: 0.20, cache_write: 2.50, cache_write_1h: 4.0 },
   'claude-sonnet-5':    { input: 2.0,  output: 10.0, cache_read: 0.20, cache_write: 2.50, cache_write_1h: 4.0 },
   'claude-sonnet-4-6':  { input: 3.0,  output: 15.0, cache_read: 0.3,  cache_write: 3.75, cache_write_1h: 6.0 },
   // Sonnet 4.5 (dated id claude-sonnet-4-5-20250929): $3/$15/$0.30/$3.75 (R7).
@@ -105,10 +111,14 @@ const DEFAULT_RATES = {
   // as plain input. Until 2026-09-22 this block said "no separate write charge" for
   // every row, which had stopped being true — Terra and Luna writes were a fifth
   // low (R27). Longest-prefix match lets each tier key win over a shorter one.
-  // GPT-6: Astra is the flagship; Sol and Luna were released 2026-09-22. Codex's
-  // config docs use gpt-6-sol as the default model and gpt-6-luna for lower-cost
-  // subagents (learn.chatgpt.com/docs/config-file/config-basic, /agent-configuration/subagents).
+  // GPT-6: Astra is the flagship; Sol and Luna were released 2026-09-22. GPT-6.1 Sol
+  // is Codex's default model from 0.159.1 (2026-09-29): `model = "gpt-6.1-sol"` in
+  // config-basic, and the subagent docs' starting model for demanding agents, with
+  // gpt-6-luna for lower-cost ones (learn.chatgpt.com/docs/config-file/config-basic,
+  // /agent-configuration/subagents, both read 2026-10-03). Its row was read from the
+  // pricing page on 2026-10-03: GPT-6 Sol's prices, except cached input at 0.05×.
   'gpt-6-astra':        { input: 10.0, output: 50.0, cache_read: 1.0,  cache_write: 12.5 },
+  'gpt-6.1-sol':        { input: 2.0,  output: 10.0, cache_read: 0.1,  cache_write: 2.5 },
   'gpt-6-sol':          { input: 2.0,  output: 10.0, cache_read: 0.2,  cache_write: 2.5 },
   'gpt-6-luna':         { input: 0.10, output: 0.50, cache_read: 0.01, cache_write: 0.125 },
   // GPT-5.6. The bare `gpt-5.6` id is an alias that "routes requests to GPT-5.6 Sol"
@@ -469,8 +479,8 @@ function readRecords(cwd) {
  */
 // No single subagent runs for six hours — the longest native-workflow phase runs
 // measured in the harness finish inside an hour — while a parent-transcript slice
-// that spans a working day, or the idle night between two stops, does. Mirrored by
-// the hooks' SLICE_MAX_DURATION_MS, which nulls the span on write. Calibrated on
+// that spans a working day, or the idle night between two stops, does. The hooks
+// write the span as measured; only this reader applies the rule. Calibrated on
 // eleven field ledgers (2026-09): of the timed rows the old ratio rule flagged,
 // 55 spanned under three hours and 12 spanned six to twenty-four — the latter all
 // parent slices booked to a `general-purpose` or workflow subagent.