first-tree 0.0.2 → 0.0.3

package/skills/first-tree/engine/validators/nodes.ts

@@ -0,0 +1,514 @@
+import { existsSync, readdirSync, readFileSync, statSync } from "node:fs";
+import { join, relative, posix } from "node:path";
+
+const FRONTMATTER_RE = /^---\s*\n(.*?)\n---/s;
+const OWNERS_RE = /^owners:\s*\[([^\]]*)\]/m;
+const SOFT_LINKS_INLINE_RE = /^soft_links:\s*\[([^\]]*)\]/m;
+const SOFT_LINKS_BLOCK_RE = /^soft_links:\s*\n((?:\s+-\s+.+\n?)+)/m;
+const TITLE_RE = /^title:\s*['"]?(.+?)['"]?\s*$/m;
+const GITHUB_USER_RE = /^[a-zA-Z0-9](?:[a-zA-Z0-9-]{0,37}[a-zA-Z0-9])?$/;
+const MD_LINK_RE = /\[.*?\]\(([^)]+\.md)\)/g;
+const DOMAIN_LINK_RE = /\[(\w[\w-]*)\/?\]\((\w[\w-]*)\/NODE\.md\)/g;
+
+const SKIP = new Set(["node_modules", "__pycache__"]);
+const SKIP_FILES = new Set(["AGENT.md", "CLAUDE.md"]);
+const MIN_BODY_LENGTH = 20;
+
+export class Findings {
+  errors: string[] = [];
+  warnings: string[] = [];
+  infos: string[] = [];
+
+  error(msg: string): void {
+    this.errors.push(msg);
+  }
+  warning(msg: string): void {
+    this.warnings.push(msg);
+  }
+  info(msg: string): void {
+    this.infos.push(msg);
+  }
+  hasErrors(): boolean {
+    return this.errors.length > 0;
+  }
+  printReport(totalFiles: number): void {
+    const all: [string, string][] = [
+      ...this.errors.map((e): [string, string] => ["error", e]),
+      ...this.warnings.map((w): [string, string] => ["warning", w]),
+      ...this.infos.map((i): [string, string] => ["info", i]),
+    ];
+    if (all.length > 0) {
+      const counts: string[] = [];
+      if (this.errors.length) counts.push(`${this.errors.length} error(s)`);
+      if (this.warnings.length) counts.push(`${this.warnings.length} warning(s)`);
+      if (this.infos.length) counts.push(`${this.infos.length} info(s)`);
+      console.log(`Found ${counts.join(", ")}:\n`);
+      const icons: Record<string, string> = {
+        error: "\u2717",
+        warning: "\u26a0",
+        info: "\u2139",
+      };
+      for (const [severity, msg] of all) {
+        console.log(`  ${icons[severity]} [${severity}] ${msg}`);
+      }
+    } else {
+      console.log(`All ${totalFiles} node(s) passed validation.`);
+    }
+  }
+}
+
+// -- Utilities --
+
+let treeRoot = "";
+const textCache = new Map<string, string | null>();
+
+export function setTreeRoot(root: string): void {
+  treeRoot = root;
+  textCache.clear();
+}
+
+export function getTreeRoot(): string {
+  return treeRoot;
+}
+
+function rel(path: string): string {
+  return relative(treeRoot, path);
+}
+
+function shouldSkip(path: string): boolean {
+  const parts = relative(treeRoot, path).split("/");
+  return parts.some((part) => SKIP.has(part) || part.startsWith("."));
+}
+
+function readText(path: string): string | null {
+  if (!textCache.has(path)) {
+    try {
+      textCache.set(path, readFileSync(path, "utf-8"));
+    } catch {
+      textCache.set(path, null);
+    }
+  }
+  return textCache.get(path)!;
+}
+
+export function parseFrontmatter(path: string): string | null {
+  const text = readText(path);
+  if (text === null) return null;
+  const m = text.match(FRONTMATTER_RE);
+  return m ? m[1] : null;
+}
+
+export function parseBody(path: string): string | null {
+  const text = readText(path);
+  if (text === null) return null;
+  const m = text.match(FRONTMATTER_RE);
+  if (m) return text.slice(m[0].length);
+  return text;
+}
+
+export function parseSoftLinks(fm: string): string[] | null {
+  // Inline format
+  let m = fm.match(SOFT_LINKS_INLINE_RE);
+  if (m) {
+    const raw = m[1].trim();
+    if (!raw) return [];
+    return raw
+      .split(",")
+      .map((s) => s.trim().replace(/^['"]|['"]$/g, ""))
+      .filter(Boolean);
+  }
+  // Block format
+  m = fm.match(SOFT_LINKS_BLOCK_RE);
+  if (m) {
+    return m[1]
+      .trim()
+      .split("\n")
+      .map((line) =>
+        line
+          .trim()
+          .replace(/^-\s*/, "")
+          .trim()
+          .replace(/^['"]|['"]$/g, ""),
+      );
+  }
+  return null;
+}
+
+function resolveSoftLink(link: string): boolean {
+  const clean = link.replace(/^\/+/, "");
+  const target = join(treeRoot, clean);
+
+  // Direct .md file
+  try {
+    if (statSync(target).isFile() && target.endsWith(".md")) return true;
+  } catch {
+    // not found
+  }
+  // Directory with NODE.md
+  try {
+    if (statSync(target).isDirectory() && existsSync(join(target, "NODE.md")))
+      return true;
+  } catch {
+    // not found
+  }
+  return false;
+}
+
+function normalizeSoftLink(link: string): string {
+  const clean = link.replace(/^\/+/, "");
+  const target = join(treeRoot, clean);
+  try {
+    if (statSync(target).isDirectory()) return join(target, "NODE.md");
+  } catch {
+    // not a directory
+  }
+  return target;
+}
+
+export function collectMdFiles(): string[] {
+  const files: string[] = [];
+  function walk(dir: string): void {
+    let entries: string[];
+    try {
+      entries = readdirSync(dir).sort();
+    } catch {
+      return;
+    }
+    for (const entry of entries) {
+      const full = join(dir, entry);
+      if (shouldSkip(full)) continue;
+      try {
+        const stat = statSync(full);
+        if (stat.isDirectory()) {
+          walk(full);
+        } else if (
+          stat.isFile() &&
+          entry.endsWith(".md") &&
+          !SKIP_FILES.has(entry)
+        ) {
+          files.push(full);
+        }
+      } catch {
+        // skip
+      }
+    }
+  }
+  walk(treeRoot);
+  return files;
+}
+
+// -- Validation checks --
+
+export function validateOwners(
+  fm: string,
+  path: string,
+  findings: Findings,
+): void {
+  const m = fm.match(OWNERS_RE);
+  if (!m) {
+    findings.error(`${rel(path)}: missing 'owners' field in frontmatter`);
+    return;
+  }
+  const raw = m[1].trim();
+  if (!raw) return; // owners: [] is valid (inheritance)
+
+  const owners = raw
+    .split(",")
+    .map((o) => o.trim())
+    .filter(Boolean);
+  if (owners.length === 0) {
+    findings.error(`${rel(path)}: owners list contains only whitespace entries`);
+    return;
+  }
+  if (owners.length === 1 && owners[0] === "*") return; // owners: [*] valid
+
+  for (const owner of owners) {
+    if (owner === "*") {
+      findings.error(
+        `${rel(path)}: wildcard '*' must be the sole entry, not mixed with usernames`,
+      );
+    } else if (!GITHUB_USER_RE.test(owner)) {
+      findings.error(`${rel(path)}: invalid owner '${owner}'`);
+    }
+  }
+}
+
+export function validateSoftLinks(
+  fm: string,
+  path: string,
+  findings: Findings,
+): void {
+  const links = parseSoftLinks(fm);
+  if (links === null) return;
+  for (const link of links) {
+    if (!link) {
+      findings.error(`${rel(path)}: empty soft_link entry`);
+    } else if (!resolveSoftLink(link)) {
+      findings.error(
+        `${rel(path)}: soft_link '${link}' does not resolve to an existing node`,
+      );
+    }
+  }
+}
+
+export function validateFolders(findings: Findings): void {
+  function walk(dir: string): void {
+    let entries: string[];
+    try {
+      entries = readdirSync(dir).sort();
+    } catch {
+      return;
+    }
+    for (const entry of entries) {
+      const full = join(dir, entry);
+      if (shouldSkip(full)) continue;
+      try {
+        if (!statSync(full).isDirectory()) continue;
+      } catch {
+        continue;
+      }
+      if (!existsSync(join(full, "NODE.md"))) {
+        findings.error(`${rel(full)}/: missing NODE.md`);
+      }
+      walk(full);
+    }
+  }
+  walk(treeRoot);
+}
+
+export function validateDirectoryListing(findings: Findings): void {
+  function walk(dir: string): void {
+    let entries: string[];
+    try {
+      entries = readdirSync(dir).sort();
+    } catch {
+      return;
+    }
+    for (const entry of entries) {
+      const full = join(dir, entry);
+      if (shouldSkip(full)) continue;
+      try {
+        if (!statSync(full).isDirectory()) continue;
+      } catch {
+        continue;
+      }
+      const nodeMd = join(full, "NODE.md");
+      if (!existsSync(nodeMd)) {
+        walk(full);
+        continue;
+      }
+      const body = parseBody(nodeMd);
+      if (body !== null) {
+        const actualLeaves = new Set<string>();
+        for (const f of readdirSync(full).sort()) {
+          const fp = join(full, f);
+          try {
+            if (statSync(fp).isFile() && f.endsWith(".md") && f !== "NODE.md") {
+              actualLeaves.add(f);
+            }
+          } catch {
+            // skip
+          }
+        }
+        const referenced = new Set<string>();
+        let linkMatch: RegExpExecArray | null;
+        const linkRe = /\[.*?\]\(([^)]+\.md)\)/g;
+        while ((linkMatch = linkRe.exec(body)) !== null) {
+          const ref = linkMatch[1];
+          if (ref.startsWith("http") || ref.startsWith("/")) continue;
+          if (!ref.includes("/")) referenced.add(ref);
+        }
+        for (const orphan of [...actualLeaves].filter((f) => !referenced.has(f)).sort()) {
+          findings.warning(
+            `${rel(nodeMd)}: leaf file '${orphan}' exists but is not mentioned in NODE.md`,
+          );
+        }
+        for (const ref of [...referenced].filter((f) => !actualLeaves.has(f)).sort()) {
+          if (!existsSync(join(full, ref))) {
+            findings.warning(
+              `${rel(nodeMd)}: references '${ref}' but the file does not exist`,
+            );
+          }
+        }
+      }
+      walk(full);
+    }
+  }
+  walk(treeRoot);
+}
+
+export function validateRootDomainSync(findings: Findings): void {
+  const nodeMd = join(treeRoot, "NODE.md");
+  const body = parseBody(nodeMd);
+  if (body === null) return;
+
+  // Strip HTML comments
+  const bodyNoComments = body.replace(/<!--.*?-->/gs, "");
+
+  const listedDomains = new Set<string>();
+  let dm: RegExpExecArray | null;
+  const domainRe = /\[(\w[\w-]*)\/?\]\((\w[\w-]*)\/NODE\.md\)/g;
+  while ((dm = domainRe.exec(bodyNoComments)) !== null) {
+    listedDomains.add(dm[2]);
+  }
+
+  const actualDomains = new Set<string>();
+  for (const child of readdirSync(treeRoot).sort()) {
+    const full = join(treeRoot, child);
+    try {
+      if (!statSync(full).isDirectory()) continue;
+    } catch {
+      continue;
+    }
+    if (child.startsWith(".") || SKIP.has(child)) continue;
+    if (existsSync(join(full, "NODE.md"))) actualDomains.add(child);
+  }
+
+  for (const missing of [...actualDomains].filter((d) => !listedDomains.has(d)).sort()) {
+    findings.error(
+      `NODE.md: domain directory '${missing}/' exists but is not listed in root NODE.md`,
+    );
+  }
+  for (const extra of [...listedDomains].filter((d) => !actualDomains.has(d)).sort()) {
+    findings.error(
+      `NODE.md: lists domain '${extra}/' but the directory does not exist or has no NODE.md`,
+    );
+  }
+}
+
+export function validateSoftLinkReciprocity(
+  files: string[],
+  findings: Findings,
+): void {
+  const allLinks: [string, string][] = [];
+
+  for (const path of files) {
+    const fm = parseFrontmatter(path);
+    if (fm === null) continue;
+    const links = parseSoftLinks(fm);
+    if (!links) continue;
+    for (const link of links) {
+      if (!link) continue;
+      const target = normalizeSoftLink(link);
+      allLinks.push([path, target]);
+    }
+  }
+
+  for (const [source, target] of allLinks) {
+    if (!existsSync(target)) continue;
+
+    let hasBackLink = false;
+    const targetFm = parseFrontmatter(target);
+    if (targetFm) {
+      const targetLinks = parseSoftLinks(targetFm);
+      if (targetLinks) {
+        for (const tl of targetLinks) {
+          if (!tl) continue;
+          const resolved = normalizeSoftLink(tl);
+          if (
+            resolved === source ||
+            resolved === join(source, "..", "NODE.md")
+          ) {
+            hasBackLink = true;
+            break;
+          }
+        }
+      }
+    }
+
+    if (!hasBackLink) {
+      const targetBody = parseBody(target);
+      if (targetBody) {
+        const sourceRel = rel(source);
+        const linkRe = /\[.*?\]\(([^)]+\.md)\)/g;
+        let lm: RegExpExecArray | null;
+        while ((lm = linkRe.exec(targetBody)) !== null) {
+          const ref = lm[1];
+          if (ref.startsWith("http") || ref.startsWith("/")) continue;
+          if (sourceRel.endsWith(ref) || ref === sourceRel) {
+            hasBackLink = true;
+            break;
+          }
+        }
+      }
+    }
+
+    if (!hasBackLink) {
+      findings.info(
+        `${rel(source)}: soft_link to '${rel(target)}' is one-way (target has no reference back)`,
+      );
+    }
+  }
+}
+
+export function validateEmptyNodes(
+  files: string[],
+  findings: Findings,
+): void {
+  for (const path of files) {
+    const text = readText(path);
+    if (text === null) continue;
+    const m = text.match(FRONTMATTER_RE);
+    if (!m) continue;
+    const body = text.slice(m[0].length);
+    const stripped = body.replace(/\s+/g, "");
+    if (stripped.length < MIN_BODY_LENGTH) {
+      findings.warning(`${rel(path)}: node has little or no body content`);
+    }
+  }
+}
+
+export function validateTitleMismatch(
+  files: string[],
+  findings: Findings,
+): void {
+  for (const path of files) {
+    const text = readText(path);
+    if (text === null) continue;
+    const fmMatch = text.match(FRONTMATTER_RE);
+    if (!fmMatch) continue;
+
+    const fm = fmMatch[1];
+    const titleMatch = fm.match(TITLE_RE);
+    if (!titleMatch) continue;
+    const fmTitle = titleMatch[1].trim();
+
+    const body = text.slice(fmMatch[0].length);
+    const headingMatch = body.match(/^#\s+(.+)$/m);
+    if (!headingMatch) continue;
+    const bodyHeading = headingMatch[1].trim();
+
+    if (fmTitle !== bodyHeading) {
+      findings.warning(
+        `${rel(path)}: frontmatter title '${fmTitle}' differs from first heading '${bodyHeading}'`,
+      );
+    }
+  }
+}
+
+export function runValidateNodes(root: string): { exitCode: number; findings: Findings } {
+  setTreeRoot(root);
+  const files = collectMdFiles();
+  const findings = new Findings();
+
+  validateFolders(findings);
+
+  for (const path of files) {
+    const fm = parseFrontmatter(path);
+    if (fm === null) {
+      findings.error(`${rel(path)}: no frontmatter found`);
+      continue;
+    }
+    validateOwners(fm, path, findings);
+    validateSoftLinks(fm, path, findings);
+  }
+
+  validateDirectoryListing(findings);
+  validateRootDomainSync(findings);
+  validateSoftLinkReciprocity(files, findings);
+  validateEmptyNodes(files, findings);
+  validateTitleMismatch(files, findings);
+
+  findings.printReport(files.length);
+  return { exitCode: findings.hasErrors() ? 1 : 0, findings };
+}

package/skills/first-tree/engine/verify.ts

@@ -0,0 +1,97 @@
+import { Repo } from "#skill/engine/repo.js";
+import { runValidateMembers } from "#skill/engine/validators/members.js";
+import { runValidateNodes } from "#skill/engine/validators/nodes.js";
+
+const UNCHECKED_RE = /^- \[ \] (.+)$/gm;
+
+export function check(label: string, passed: boolean): boolean {
+  const icon = passed ? "\u2713" : "\u2717";
+  const status = passed ? "PASS" : "FAIL";
+  console.log(`  ${icon} [${status}] ${label}`);
+  return passed;
+}
+
+export function checkProgress(repo: Repo): string[] {
+  const progressPath = repo.progressPath();
+  const text = progressPath === null ? null : repo.readFile(progressPath);
+  if (text === null) return [];
+  const matches: string[] = [];
+  let m: RegExpExecArray | null;
+  UNCHECKED_RE.lastIndex = 0;
+  while ((m = UNCHECKED_RE.exec(text)) !== null) {
+    matches.push(m[1]);
+  }
+  return matches;
+}
+
+export interface ValidateNodesResult {
+  exitCode: number;
+}
+
+export type NodeValidator = (root: string) => ValidateNodesResult;
+
+function defaultNodeValidator(root: string): ValidateNodesResult {
+  const { exitCode } = runValidateNodes(root);
+  return { exitCode };
+}
+
+export function runVerify(repo?: Repo, nodeValidator?: NodeValidator): number {
+  const r = repo ?? new Repo();
+  const validate = nodeValidator ?? defaultNodeValidator;
+  let allPassed = true;
+  const progressPath = r.progressPath() ?? r.preferredProgressPath();
+  const frameworkVersionPath = r.frameworkVersionPath();
+
+  console.log("Context Tree Verification\n");
+
+  // Progress file check
+  const unchecked = checkProgress(r);
+  if (unchecked.length > 0) {
+    console.log(`  Unchecked items in ${progressPath}:\n`);
+    for (const item of unchecked) {
+      console.log(`    - [ ] ${item}`);
+    }
+    console.log();
+    console.log(
+      `  Verify each step above and check it off in ${progressPath} before running verify again.\n`,
+    );
+    allPassed = false;
+  }
+
+  // Deterministic checks
+  console.log("  Checks:\n");
+
+  // 1. Framework exists
+  allPassed = check(`${frameworkVersionPath} exists`, r.hasFramework()) && allPassed;
+
+  // 2. Root NODE.md has valid frontmatter
+  const fm = r.frontmatter("NODE.md");
+  const hasValidNode =
+    fm !== null && fm.title !== undefined && fm.owners !== undefined;
+  allPassed = check(
+    "Root NODE.md has valid frontmatter (title, owners)",
+    hasValidNode,
+  ) && allPassed;
+
+  // 3. AGENT.md exists with framework markers
+  allPassed = check(
+    "AGENT.md exists with framework markers",
+    r.hasAgentMdMarkers(),
+  ) && allPassed;
+
+  // 4. Node validation
+  const { exitCode } = validate(r.root);
+  allPassed = check("Node validation passes", exitCode === 0) && allPassed;
+
+  // 5. Member validation
+  const members = runValidateMembers(r.root);
+  allPassed = check("Member validation passes", members.exitCode === 0) && allPassed;
+
+  console.log();
+  if (allPassed) {
+    console.log("All checks passed.");
+  } else {
+    console.log("Some checks failed. See above for details.");
+  }
+  return allPassed ? 0 : 1;
+}

package/skills/first-tree/references/about.md

@@ -0,0 +1,36 @@
+---
+title: "About Context Tree"
+owners: []
+---
+
+# About Context Tree
+
+**context-tree.ai** — The living source of truth for your organization.
+
+---
+
+## The Problem
+
+Organizations generate decisions constantly — in PRs, meetings, Slack threads, documents. But none of these systems stay current. PRs get merged, issues get closed, documents decay. The knowledge that produced them scatters and disappears.
+
+When an agent — or a new teammate — needs to understand *why* something was built a certain way, there's nowhere to look. The information existed once, but no system kept it alive.
+
+---
+
+## The Idea
+
+Context Tree is the **living source of truth** for an organization — a tree-structured knowledge base that agents and humans build and maintain together.
+
+Every node represents a domain, decision, or design. Every node has an **owner**. When a decision is made, it is written to the tree. When things change, the tree updates. The tree is never a snapshot — it's the current state.
+
+The result is an organization where:
+
+- Every agent and every human reads from the same, always-current source
+- Decisions are traceable — the *what*, the *why*, and who owns it
+- Knowledge compounds over time instead of evaporating
+
+---
+
+## Who It's For
+
+Agent-centric teams — founders, engineers, and product builders who work alongside AI agents every day and want their organizational knowledge to grow with them, not against them.

package/skills/first-tree/references/maintainer-architecture.md

@@ -0,0 +1,59 @@
+# Maintainer Architecture
+
+This reference explains how to maintain the `first-tree` source repo itself.
+
+## What This Repo Ships
+
+- One canonical skill: `skills/first-tree/`
+- One thin CLI package: the `context-tree` command distributed by the `first-tree`
+  npm package
+- The published package carries that canonical skill directly; normal install
+  and upgrade flows should not depend on cloning this source repo
+
+This repo is not a user context tree. User decision content lives in the repos
+that install the framework.
+
+## Canonical Layers
+
+1. `SKILL.md` defines when to use the skill and the maintainer workflow.
+2. `references/` stores the knowledge an agent needs to maintain the framework
+   and the thin CLI without reading repo-local prose.
+3. `assets/framework/` stores the runtime payload that gets installed into user
+   repos.
+4. `engine/` stores the canonical framework and CLI behavior.
+5. `tests/` store the canonical skill validation surface.
+6. The root repo may also keep maintainer-only developer tooling such as
+   `evals/` when that tooling should not ship with the skill.
+7. The root CLI/package files are implementation shell code. They should call
+   into the skill-owned engine and validation surface, not become a second
+   source of framework knowledge.
+
+## Non-Negotiables
+
+- Treat `skills/first-tree/` as the only canonical source.
+- If a maintainer needs information to safely change behavior, move that
+  information into `references/`; do not leave it only in root `README.md`,
+  `AGENT.md`, CI comments, or PR descriptions.
+- Keep runtime assets generic. They are copied into every user tree.
+- Keep the CLI thin. Command semantics, upgrade rules, layout contracts, and
+  maintainer guidance should belong to the skill.
+- Keep the user tree decision-focused. Execution detail stays in source systems.
+
+## Change Discipline
+
+- Path or layout changes: update `references/upgrade-contract.md`, task text,
+  validators, and tests together.
+- Shipped payload changes: update `assets/framework/`, the maintainer references
+  that describe the contract, and the validation surface together.
+- Thin shell changes: update the relevant maintainer reference before or during
+  the code change so the skill remains self-sufficient.
+
+## End-State Target
+
+- skill owns knowledge, runtime payload, framework engine, and the canonical
+  framework test surface
+- root owns only the light CLI/bootstrap/build shell plus maintainer-only
+  developer tooling such as `evals/`
+
+When deciding where a new file should live, bias toward the skill unless the
+file is purely package-tooling shell code.