delimit-cli 3.15.8 → 3.15.10

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -1,5 +1,46 @@
1
1
  # Changelog
2
2
 
3
+ ## [3.15.9] - 2026-03-30
4
+
5
+ ### Added
6
+ - **Agent Swarm**: 20 agents across 4 ventures with namespace isolation, tiered approvals, and central governor
7
+ - **Prompt Playbook**: versioned, reusable prompt templates with {{variable}} syntax
8
+ - **Multi-model code review**: consolidated feedback from Claude, Grok, Gemini, Codex
9
+ - **PII/secret redaction**: scan and redact API keys, emails, SSNs before sending to LLMs
10
+ - **Collision detection**: prevent two AI models from editing the same file
11
+ - **Prompt drift detection**: track when same task behaves differently across models
12
+ - **Cost vs Efficacy dashboard**: which model is best for your codebase
13
+ - **Project config**: committable `delimit.yml` with per-repo AI governance
14
+ - **`delimit resume`**: show what you were working on last session
15
+ - **`delimit quickstart`**: clone demo project and guided walkthrough
16
+ - **`delimit try`**: zero-risk demo that saves a Markdown report
17
+ - **Change management policy**: docs freshness check before deploy
18
+ - **GTM metrics tracking**: tasks, deploys, revenue per venture
19
+
20
+ ### Changed
21
+ - Hero messaging: "Stop Re-Explaining Your Codebase" (pain-first)
22
+ - Progressive disclosure: 5 workflows instead of 162 tools
23
+ - Deliberation engine now uses official SDKs (anthropic, openai, google-genai)
24
+ - 4-model deliberation: Claude (CLI) + Grok + Gemini + Codex
25
+ - Setup auto-updates, regenerates shims, fixes config permissions
26
+ - Boot screen shows dynamic version and tool count with gradient colors
27
+ - All confirmation prompts show "Enter = Yes" hint
28
+ - Pricing page updated: Free/Pro/Enterprise tiers
29
+
30
+ ### Fixed
31
+ - 37 missing @mcp.tool() decorators restored (125 to 171 tools)
32
+ - Stale .so binaries shadowing updated .py source files
33
+ - Codex config.toml duplicate TOML entries
34
+ - CI green across Python 3.10/3.11/3.12 (7 test files fixed)
35
+ - Setup self-update infinite loop guard
36
+ - Pro module stubs overwriting full source files
37
+
38
+ ### Security
39
+ - 7 proprietary modules stubbed in npm bundle
40
+ - Deliberation engine kept private (stub in npm, full on server)
41
+ - PII scrubbed from all public packages
42
+ - Security check runs before every npm publish
43
+
3
44
  ## [3.13.3] - 2026-03-27
4
45
 
5
46
  ### Changed
package/README.md CHANGED
@@ -50,15 +50,16 @@ npx delimit-cli init # Sets up governance + drift baseline
50
50
 
51
51
  ---
52
52
 
53
- ## What's New in v3.14
54
-
55
- - **Evidence timeline + gate status dashboard** -- see every governance decision and when it happened
56
- - **Multi-agent orchestration** -- track which AI assistant works on what, across models
57
- - **Continuous drift monitoring** -- detect spec drift between deployments automatically
58
- - **PR governance copilot** -- gate status posted directly in PR comments
59
- - **Compliance templates** -- SOC2, PCI-DSS, and HIPAA presets out of the box
60
- - **Beta capture in CLI** -- opt into early features from the command line
61
- - **112+ MCP tools** -- governance, context, shipping, observability, and orchestration
53
+ ## What's New in v3.15
54
+
55
+ - **Agent Swarm** -- 20 AI agents across 4 ventures with namespace isolation and tiered approvals
56
+ - **Prompt Playbook** -- versioned, reusable prompts with {{variable}} syntax
57
+ - **Multi-model code review** -- consolidated feedback from 3+ AI models
58
+ - **PII/secret redaction** -- auto-scan and redact before sending to external LLMs
59
+ - **Collision detection** -- prevent two AI models from editing the same file
60
+ - **4-model deliberation** -- Claude + Grok + Gemini + Codex debate until consensus
61
+ - **Change management** -- docs freshness check enforced before deploy
62
+ - **171 MCP tools** -- governance, context, shipping, observability, orchestration, and swarm
62
63
 
63
64
  ---
64
65
 
@@ -3692,6 +3692,11 @@ def delimit_swarm(action: str = "status", venture: str = "",
3692
3692
  if action == "metrics":
3693
3693
  from ai.swarm import get_metrics
3694
3694
  return _with_next_steps("swarm", _safe_call(get_metrics, venture=venture))
3695
+ if action == "docs_check":
3696
+ from ai.swarm import check_docs_freshness
3697
+ return _with_next_steps("swarm", _safe_call(
3698
+ check_docs_freshness, project_path=repo_path or ".",
3699
+ ))
3695
3700
  return _with_next_steps("swarm", _safe_call(get_swarm_status))
3696
3701
 
3697
3702
 
@@ -493,3 +493,100 @@ def get_metrics(venture: str = "", days: int = 7) -> Dict[str, Any]:
493
493
  "venture_filter": venture or "all",
494
494
  "days": days,
495
495
  }
496
+
497
+
498
+ # ═══════════════════════════════════════════════════════════════════════
499
+ # Change Management — docs freshness check before deploy
500
+ # ═══════════════════════════════════════════════════════════════════════
501
+
502
+ DOCS_CHECKLIST = [
503
+ {"file": "README.md", "check": "tool_count", "pattern": r"\d+ (?:MCP |governance )?tools"},
504
+ {"file": "README.md", "check": "version_badge", "pattern": r"GitHub%20Action-v[\d.]+"},
505
+ {"file": "README.md", "check": "cli_commands", "pattern": r"npx delimit-cli"},
506
+ ]
507
+
508
+
509
+ def check_docs_freshness(
510
+ project_path: str = ".",
511
+ tool_count: int = 0,
512
+ version: str = "",
513
+ ) -> Dict[str, Any]:
514
+ """Check if documentation is up-to-date before deploying.
515
+
516
+ Verifies README, changelog, and landing page reflect current
517
+ tool count, version, and feature set.
518
+ """
519
+ import re
520
+ p = Path(project_path).resolve()
521
+ findings = []
522
+ stale = False
523
+
524
+ # Check README exists
525
+ readme = p / "README.md"
526
+ if not readme.exists():
527
+ findings.append({"file": "README.md", "status": "missing", "severity": "warning"})
528
+ else:
529
+ content = readme.read_text()
530
+
531
+ # Check tool count
532
+ if tool_count > 0:
533
+ counts = re.findall(r'(\d+)\s*(?:MCP |governance )?tools', content)
534
+ for count_str in counts:
535
+ count = int(count_str)
536
+ if abs(count - tool_count) > 10:
537
+ findings.append({
538
+ "file": "README.md",
539
+ "status": "stale",
540
+ "issue": f"Says {count} tools, actual is {tool_count}",
541
+ "severity": "warning",
542
+ })
543
+ stale = True
544
+
545
+ # Check version badge
546
+ if version:
547
+ if version not in content:
548
+ findings.append({
549
+ "file": "README.md",
550
+ "status": "stale",
551
+ "issue": f"Version badge doesn't show {version}",
552
+ "severity": "info",
553
+ })
554
+
555
+ # Check What's New / CHANGELOG
556
+ changelog = p / "CHANGELOG.md"
557
+ if changelog.exists():
558
+ cl_content = changelog.read_text()
559
+ if version and version not in cl_content:
560
+ findings.append({
561
+ "file": "CHANGELOG.md",
562
+ "status": "stale",
563
+ "issue": f"No entry for version {version}",
564
+ "severity": "warning",
565
+ })
566
+ stale = True
567
+
568
+ # Check for uncommitted changes
569
+ try:
570
+ import subprocess
571
+ result = subprocess.run(
572
+ ["git", "status", "--porcelain"], capture_output=True, text=True,
573
+ cwd=str(p), timeout=5,
574
+ )
575
+ uncommitted = len(result.stdout.strip().split("\n")) if result.stdout.strip() else 0
576
+ if uncommitted > 0:
577
+ findings.append({
578
+ "file": "working tree",
579
+ "status": "dirty",
580
+ "issue": f"{uncommitted} uncommitted file(s)",
581
+ "severity": "info",
582
+ })
583
+ except Exception:
584
+ pass
585
+
586
+ return {
587
+ "status": "stale" if stale else "fresh",
588
+ "findings": findings,
589
+ "total_issues": len(findings),
590
+ "stale": stale,
591
+ "message": f"{len(findings)} doc issue(s) found" if findings else "Documentation is up to date",
592
+ }
@@ -42,6 +42,13 @@ The founder reviews and approves via email. The AI's job is to draft, send, and
42
42
  - Deploy requested: \`delimit_security_audit\` then \`delimit_test_smoke\` then \`delimit_changelog\` then \`delimit_deploy_plan\`
43
43
  - Deploy succeeds: \`delimit_deploy_verify\` + \`delimit_evidence_collect\`
44
44
  - Verify succeeds: \`delimit_ledger_done\` (related task) + \`delimit_notify\`
45
+
46
+ ### Change Management (update docs with every ship)
47
+ - After shipping a feature: update README if CLI commands or tool count changed
48
+ - After version bump: update version badge, tool count, test count on landing page
49
+ - After npm publish: sync gateway server.py to npm bundle, scrub PII, run security-check.sh
50
+ - After changing pricing/tiers: update pricing page and Lemon Squeezy product
51
+ - Before deploy: call \`delimit_swarm(action="docs_check")\` to verify docs freshness
45
52
  - Before submitting a PR to an external repo: call \`delimit_deliberate\` on the diff + PR description (Pro only — if using free tier deliberations and quota is exhausted, skip this gate but log a warning)
46
53
 
47
54
  ### Content / Outreach (all drafts, never auto-post)
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "delimit-cli",
3
3
  "mcpName": "io.github.delimit-ai/delimit-mcp-server",
4
- "version": "3.15.8",
4
+ "version": "3.15.10",
5
5
  "description": "Unify Claude Code, Codex, Cursor, and Gemini CLI with persistent context, governance, and multi-model debate.",
6
6
  "main": "index.js",
7
7
  "files": [