company-skill 1.0.0 → 1.2.0

package/README.md

@@ -35,7 +35,7 @@ GOAL: "Build the auth system"
   Done? Write STATUS.md, report to user.
 ```
 
-No arbitrary limit. The loop runs until the objective is verified done.
+No arbitrary limit. The loop runs until ALL criteria in `criteria.json` pass. A Stop Hook blocks Claude from exiting early. To cancel: `touch .company/CANCEL`.
 
 ## Built-In Roles
 
@@ -120,7 +120,7 @@ Employees show with colors: leads (cyan), workers (green), reviewers (yellow), d
 
 | Agent | Phase | Color | Role |
 |-------|-------|-------|------|
-| company-lead | THINK | Blue | Department leads, deciding what to do |
+| company-lead | THINK | Cyan | Department leads, deciding what to do |
 | company-worker | EXECUTE | Green | Employees doing the actual work |
 | company-reviewer | VERIFY | Yellow | Internal Reviewer, checking quality |
 | company-critic | VERIFY | Yellow | Devil's Advocate, finding holes |
@@ -146,8 +146,6 @@ Auto-installed on first run:
 | gstack | /review, /ship, /qa, /investigate, /browse, /office-hours |
 | GSD | /gsd:plan-phase, /gsd:execute-phase, /gsd:verify-work, /gsd:debug |
 | trailofbits | Security audit, vulnerability detection |
-| claude-mem | Persistent memory compression across sessions |
-| oh-my-claudecode | Team orchestration, 32 specialized employees, autopilot |
 
 Install manually for more:
 
@@ -155,6 +153,8 @@ Install manually for more:
 /plugin marketplace add obra/superpowers-marketplace
 /plugin marketplace add wshobson/agents
 /plugin marketplace add alirezarezvani/claude-skills
+npm i -g claude-mem
+npm i -g oh-my-claude-sisyphus
 ```
 
 When installed, employees MUST use them. Raw tools only when no skill matches the task.

package/hooks/stop-guard.js

@@ -2,50 +2,76 @@
 
 /**
  * Stop Hook Guard for /company skill.
- *
- * Blocks Claude from stopping until the goal is verified done.
+ * Blocks Claude from stopping until ALL criteria in criteria.json pass.
  * Stolen from oh-my-claudecode's persistent-mode pattern.
- *
- * Install: Add to settings.json hooks.Stop
  */
 
 const fs = require('fs');
 const path = require('path');
 
-// Check if company skill is active
+const criteriaPath = path.join(process.cwd(), '.company', 'criteria.json');
 const goalPath = path.join(process.cwd(), '.company', 'GOAL.md');
-const statusPath = path.join(process.cwd(), '.company', 'STATUS.md');
+const cancelPath = path.join(process.cwd(), '.company', 'CANCEL');
 
-if (!fs.existsSync(goalPath)) {
-  // Company not running, allow stop
+// No company running, allow stop
+if (!fs.existsSync(goalPath) && !fs.existsSync(criteriaPath)) {
   process.exit(0);
 }
 
-// Check if goal is achieved
-if (fs.existsSync(statusPath)) {
-  const status = fs.readFileSync(statusPath, 'utf8');
-  if (status.includes('ACHIEVED') || status.includes('DONE')) {
-    // Goal achieved, allow stop
+// Cancel signal, allow stop
+if (fs.existsSync(cancelPath)) {
+  try { fs.unlinkSync(cancelPath); } catch (e) {}
+  process.exit(0);
+}
+
+// Check criteria.json
+if (fs.existsSync(criteriaPath)) {
+  try {
+    const data = JSON.parse(fs.readFileSync(criteriaPath, 'utf8'));
+    const all = data.criteria || [];
+    const passing = all.filter(c => c.passes === true && c.evidence);
+    const failing = all.filter(c => !c.passes || !c.evidence);
+
+    if (all.length > 0 && failing.length === 0) {
+      // All criteria pass, allow stop
+      process.exit(0);
+    }
+
+    // Some criteria still failing, block stop
+    const failList = failing.map(c => `- ${c.description}`).join('\n');
+    const output = {
+      continue: false,
+      stopReason: `${failing.length}/${all.length} criteria not met. Continuing.`,
+      hookSpecificOutput: {
+        hookEventName: "Stop",
+        additionalContext: `COMPANY GOAL NOT ACHIEVED. ${failing.length} of ${all.length} criteria still failing:\n${failList}\n\nContinue the THINK > EXECUTE > VERIFY cycle. Read .company/criteria.json and the latest cycle briefing.`
+      }
+    };
+    console.log(JSON.stringify(output));
+    process.exit(0);
+  } catch (e) {
+    // criteria.json malformed, allow stop to prevent deadlock
     process.exit(0);
   }
 }
 
-// Check for cancel signal
-const cancelPath = path.join(process.cwd(), '.company', 'CANCEL');
-if (fs.existsSync(cancelPath)) {
-  fs.unlinkSync(cancelPath);
-  process.exit(0);
+// No criteria.json but goal exists, check STATUS.md
+const statusPath = path.join(process.cwd(), '.company', 'STATUS.md');
+if (fs.existsSync(statusPath)) {
+  const status = fs.readFileSync(statusPath, 'utf8');
+  if (status.includes('ACHIEVED')) {
+    process.exit(0);
+  }
 }
 
-// Goal not achieved, block stop and continue
-const goal = fs.readFileSync(goalPath, 'utf8');
+// Default: block stop
+const goal = fs.existsSync(goalPath) ? fs.readFileSync(goalPath, 'utf8').substring(0, 500) : 'Unknown goal';
 const output = {
   continue: false,
-  stopReason: "Company goal not yet achieved. Continuing cycles.",
+  stopReason: "Company goal not yet achieved.",
   hookSpecificOutput: {
     hookEventName: "Stop",
-    additionalContext: `The /company goal is NOT yet achieved. Continue the THINK > EXECUTE > VERIFY cycle. Read .company/GOAL.md and the latest cycle briefing to continue.\n\nGoal:\n${goal.substring(0, 500)}`
+    additionalContext: `Goal not achieved. Continue cycles.\n\n${goal}`
   }
 };
-
 console.log(JSON.stringify(output));

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "company-skill",
-  "version": "1.0.0",
+  "version": "1.2.0",
   "description": "Goal-driven multi-employee company for Claude Code. Give it a goal, it runs until done.",
   "bin": {
     "company-skill": "./bin/install.js"

package/skill/SKILL.md

@@ -146,16 +146,45 @@ mkdir -p .company/cycles .company/messages .company/memory
 grep -q "^\.company/" .gitignore 2>/dev/null || echo ".company/" >> .gitignore 2>/dev/null || true
 ```
 
-Write `.company/GOAL.md`:
+Write `.company/GOAL.md` with STRUCTURED checkable criteria (not free text):
 ```markdown
 # Goal
 {the user's goal}
 
 # Success Criteria
-{inferred from the goal, what does "done" look like?}
+Each criterion MUST be a yes/no checkable statement. No vague language.
+
+- [ ] {specific measurable criterion 1}
+- [ ] {specific measurable criterion 2}
+- [ ] {specific measurable criterion 3}
+
+Bad examples (REJECT these):
+- "Code is clean" (vague)
+- "Performance is good" (not measurable)
+- "Implementation is complete" (not checkable)
+
+Good examples:
+- "Function X returns Y when given input Z"
+- "Compression ratio > 20x measured with honest byte counting"
+- "All tests pass with 0 failures"
+- "README has install instructions that work on a fresh machine"
 ```
 
-Write `.company/cycles/cycle-0-briefing.md` with: goal, success criteria, team structure, rules, any previous state.
+Write `.company/criteria.json` (machine-checkable state):
+```json
+{
+  "goal": "{the user's goal}",
+  "criteria": [
+    {"id": 1, "description": "{criterion 1}", "passes": false, "evidence": null},
+    {"id": 2, "description": "{criterion 2}", "passes": false, "evidence": null},
+    {"id": 3, "description": "{criterion 3}", "passes": false, "evidence": null}
+  ]
+}
+```
+
+The loop ONLY exits when ALL criteria have `"passes": true` with non-null evidence.
+
+Write `.company/cycles/cycle-0-briefing.md` with: goal, criteria, team structure, rules, any previous state.
 
 ## Step 4: Run Loop (repeat until verified)
 
@@ -251,33 +280,30 @@ This is what makes the loop work. Launch TWO built-in reviewers:
 ```
 You are the Internal Reviewer. Cycle {N} just completed.
 
-GOAL: {from .company/GOAL.md}
-SUCCESS CRITERIA: {from .company/GOAL.md}
-
-Read ALL of these:
-- All .company/messages/*.jsonl from this cycle
-- All .company/{dept}/*.md employee findings
-- Any code changes (git diff if applicable)
+Read .company/criteria.json and ALL .company/messages/*.jsonl and .company/{dept}/*.md findings.
 
-QUESTION: Has the goal been achieved? Check each success criterion.
+VERIFICATION RULES:
+- For EACH criterion in criteria.json with "passes": false, check if this cycle produced evidence.
+- Every finding MUST have a SOURCE field. REJECT any finding without one.
+- For priority 4-5 findings, RE-RUN the experiment or RE-CHECK the source file yourself.
+- If a number is claimed, read the actual output file that produced it.
+- Any claim you cannot independently verify: mark UNVERIFIED.
 
-DOUBLE-VERIFICATION RULES:
-- Every finding MUST have a SOURCE field. Reject any without one.
-- For HIGH-priority findings (4-5), RE-RUN the experiment or RE-CHECK the source yourself.
-- If a finding says "24x compression", verify by reading the actual experiment output file.
-- If a finding cites a paper, verify the paper exists via WebSearch.
-- Any claim you cannot independently verify gets marked UNVERIFIED.
+UPDATE .company/criteria.json:
+- For each criterion where evidence now exists, set "passes": true, fill "evidence" with proof.
+- If no evidence, keep "passes": false.
 
 Write to .company/cycles/cycle-{N}-review.md:
 For each criterion:
-  CRITERION: {what was required}
-  STATUS: MET / NOT MET / PARTIALLY MET
-  EVIDENCE: {what proves it, with file path or URL}
-  VERIFIED: YES (I re-checked) / NO (could not verify)
-  GAPS: {what's still missing}
-
-FINAL VERDICT: DONE or NOT DONE
-If NOT DONE: list exactly what the next cycle must fix.
+  ID: {from criteria.json}
+  DESCRIPTION: {criterion}
+  PASSES: true/false
+  EVIDENCE: {file path, URL, or command output}
+  VERIFIED: YES/NO
+
+FINAL VERDICT:
+- If ALL criteria in criteria.json have "passes": true with non-null evidence: DONE
+- If ANY has "passes": false: NOT DONE, list exactly what next cycle must do
 ```
 
 **Devil's Advocate:**