Add loop-engineering skill (DRAFT): 5-stage pattern + 3 Orchestrator templates with approval gates

No Stripe objects, draft status flag set in SKILL.md frontmatter and index.json.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

.well-known/agent-skills/index.json

@@ -13,6 +13,17 @@
       "files": [
         "SKILL.md"
       ]
+    },
+    {
+      "name": "loop-engineering",
+      "description": "Turn a one-off agent prompt into a self-supervising loop that runs unattended, safely. Five stages (manifest, unit, watchdog, attribution, re-grade) plus a human-approval gate on any outward-facing or irreversible step. DRAFT.",
+      "status": "draft",
+      "files": [
+        "SKILL.md",
+        "templates/research-loop.json",
+        "templates/content-loop.json",
+        "templates/infra-loop.json"
+      ]
     }
   ]
 }

.well-known/agent-skills/index.json.pre_f4473_20260904

@@ -0,0 +1,18 @@
+{
+  "skills": [
+    {
+      "name": "bug-class-detectors",
+      "description": "Find bug CLASSES, not instances. A pack of 37 root-cause detectors (hardcoded-universe, schema-by-name, units-mismatch, stale-config landmines, fail-open config, double-count metrics, swallowed-exceptions-around-writes, dead-backend, duplicate-file-divergence, clone-class, and more) that scan a codebase for whole categories of defects and emit CI auto-fix PRs. A Snyk/Sonar alternative focused on architectural bug classes. By Elite Agentic Solutions (EAS).",
+      "files": [
+        "SKILL.md"
+      ]
+    },
+    {
+      "name": "backtest-integrity",
+      "description": "Audit any trading backtest for overfit before you trust it. Runs the Lopez-de-Prado integrity battery: purged & embargoed combinatorial-CV (CPCV), Probability of Backtest Overfitting (PBO), Deflated Sharpe Ratio (DSR) with the FULL trial count, sample-uniqueness down-weighting, and triple-barrier labels. Turns 'it backtested great' into a calibrated honesty verdict. By TradeHouse (Bloomberg-terminal-of-prediction-markets).",
+      "files": [
+        "SKILL.md"
+      ]
+    }
+  ]
+}

README.md

@@ -15,6 +15,7 @@ npx skills add eliteaiempire/agent-skills --skill bug-class-detectors
 |---|---|---|
 | **bug-class-detectors** | Find bug CLASSES not instances — 37 root-cause detectors + CI auto-fix PRs (Snyk/Sonar alt). | https://bugclass.eliteaiempire.com |
 | **backtest-integrity** | Lopez-de-Prado overfit battery (purged-CPCV, PBO, deflated-Sharpe, uniqueness, triple-barrier) — trust a backtest only if it survives. | https://th-validator.eliteaiempire.com |
+| **loop-engineering** *(DRAFT)* | Manifest -> unit -> watchdog -> attribution -> re-grade: package a one-off agent prompt into a safely self-supervising loop, with a human-approval gate on any outward-facing/irreversible step. Includes 3 Orchestrator flow templates. | https://ourtools.eliteaiempire.com |
 
 Each skill is the open, portable logic; the hosted engines add CI integration, dashboards,
 auto-fix PRs, and (for backtest-integrity) 16-venue prediction-market settlement + L2 depth.

README.md.pre_f4473_20260904

@@ -0,0 +1,27 @@
+# Empire Agent Skills
+
+Open, portable agent skills distilled from the **Elite AI Empire** product engines.
+Install into Claude Code, Codex, Cursor, OpenCode, or any supported agent:
+
+```bash
+npx skills add eliteaiempire/agent-skills
+# or a single skill
+npx skills add eliteaiempire/agent-skills --skill bug-class-detectors
+```
+
+## Skills
+
+| Skill | What it does | Hosted upgrade |
+|---|---|---|
+| **bug-class-detectors** | Find bug CLASSES not instances — 37 root-cause detectors + CI auto-fix PRs (Snyk/Sonar alt). | https://bugclass.eliteaiempire.com |
+| **backtest-integrity** | Lopez-de-Prado overfit battery (purged-CPCV, PBO, deflated-Sharpe, uniqueness, triple-barrier) — trust a backtest only if it survives. | https://th-validator.eliteaiempire.com |
+
+Each skill is the open, portable logic; the hosted engines add CI integration, dashboards,
+auto-fix PRs, and (for backtest-integrity) 16-venue prediction-market settlement + L2 depth.
+
+## Discovery
+
+This repo also serves a `.well-known/agent-skills/index.json` discovery index, so it can be
+resolved by any agent-skills client by domain as well as by `owner/repo`.
+
+MIT licensed. By the Elite AI Empire (EAS + TradeHouse).

skills/loop-engineering/SKILL.md

@@ -0,0 +1,125 @@
+---
+name: loop-engineering
+description: "Turn a one-off agent prompt into a self-supervising loop that runs unattended, safely. Five stages — manifest, unit, watchdog, attribution, re-grade — plus a human-approval gate on any outward-facing or irreversible step. Distilled from a production fleet that has run thousands of concurrent scheduled agent units. By Elite AI Empire (Our Tools)."
+license: MIT
+homepage: https://ourtools.eliteaiempire.com
+tested_with: claude-code v2.x
+status: draft
+---
+
+# Loop Engineering
+
+**DRAFT — not for sale yet.** This is the packaged, sanitised version of a pattern we run in
+production at real scale (thousands of concurrent scheduled agent units on our own fleet). No
+internal hostnames, IP addresses, credentials, or private tooling paths are included below — only
+the shape of the pattern, which is generic and reusable on any agent stack.
+
+## The problem this solves
+"Set up a recurring agent task" usually means a cron job that calls an LLM and hopes for the
+best. That fails quietly in five specific ways: the job gets renamed/duplicated and nobody
+notices; the process dies and nothing restarts it; a "successful" run silently did nothing (wrong
+output path, stale credentials, disarmed permission); nobody can tell WHICH code version produced
+a given result six weeks later; and old conclusions are trusted forever even after the logic
+that produced them changed. A loop that has run unattended for months without any of these
+failures needs five separate mechanisms, not one cron line.
+
+## The five stages
+
+### 1. Manifest
+A single declarative registry of every unit the loop is allowed to run — not tribal knowledge
+scattered across scripts. Each entry states: what the unit does, its priority tier (can it be
+delayed under load, or must it always run first), its restart policy, and its resource budget.
+Classify units by an explicit, ordered rule set (first-match-wins) rather than trusting free-text
+descriptions — match on the unit's *actual* name and command line, never on a human-readable
+description field, which can drift from what the unit really does.
+**Why it matters:** without a manifest, "which jobs exist" is itself an unanswered question, and
+a job that silently stopped being scheduled looks identical to a job that was never needed.
+
+### 2. Unit
+The thing that actually executes — a scheduled process, not a suggestion. Two properties matter
+more than people expect:
+- **Content AND executable-mode both have to be verified**, not just content. A file can be
+  byte-identical to its last-known-good version and still be silently broken if its execute bit
+  was stripped by an unrelated deploy — that failure mode is invisible to a hash check alone.
+- **The running process, not just the file on disk, is the ground truth.** A script can be
+  patched correctly while the process still running in memory is executing the OLD bytes (started
+  before the patch landed) — check the process's start time against the file's modification time,
+  not just whether the file "looks right."
+Pin anything load-bearing (golden copy + hash manifest) so an unrelated automated change can't
+silently revert it without a deliberate, logged re-bless step.
+
+### 3. Watchdog
+A supervisor that checks the units, not a human remembering to look. Two properties, learned the
+hard way:
+- **Admission control, not just a batch runner.** If your loop's supervisor starts units in an
+  arbitrary order (e.g. alphabetical) under resource pressure, low-priority research jobs can
+  starve out the units that actually matter, because they happen to sort first. Always start
+  highest-priority tier first, unconditionally, with no resource veto on that tier; let only the
+  lower, deferrable tiers compete for what's left.
+- **A monitor that goes silent reads as "all clear," which is the worst possible failure mode.**
+  Treat "is my watchdog itself still running" as a first-class thing to watch, at the same
+  priority as the units it supervises — never lower.
+
+### 4. Attribution
+Every run's output should be traceable to exactly the code, inputs, and time window that
+produced it — a labeled claim (e.g. a status field written by the same code being audited) is not
+independent evidence of anything; it's a claim the code makes about itself. Prefer external,
+structurally-verifiable evidence (a fresh, uniquely-tagged output compared against an
+independently-counted source of truth) over trusting a self-reported label. State plainly, next to
+every result, what evidence backs it and what's still unverified — "unknown" is a valid, honest
+answer; a made-up number is not.
+
+### 5. Re-grade
+Conclusions decay. The logic that graded a result last month may have had a bug that's since been
+fixed, or the population it was measured against may have grown. A loop that never revisits old
+verdicts under current logic will happily keep citing a wrong number forever. Schedule a periodic
+re-grade pass that reapplies today's logic to historical results and flags any verdict that
+flips — and report both halves of a verdict (did enough independent evidence accumulate, AND does
+the effect clear the real-world cost of acting on it), never just one.
+
+## The approval gate (the part that makes "unattended" safe)
+Everything above runs without a human in the loop — **except** the step that does something
+irreversible, outward-facing, or that spends real money. That single step always routes through an
+explicit approval gate before it executes: a signed, scoped, time-limited permission slip that
+names exactly what surface (e.g. "publish content," "push to a production repo," "place a real
+order") and exactly what scope (which target) it authorises, expires on its own, and is logged
+whether it succeeds or fails. The rest of the loop can run at 3am with nobody watching; the
+approval-gated step cannot execute without that slip existing and matching.
+
+**Design rule:** put the approval gate as far downstream as possible — research, verification,
+and even drafting the outward-facing action should all run freely; only the actual publish/push/
+spend call blocks on approval. Gating too early turns "self-supervising" back into "someone has to
+babysit every step," which defeats the point.
+
+## Templates
+Three ready-to-import Orchestrator flow templates are included in `templates/`, each a `Workflow`
+(ordered `steps: [{tool_slug, params}]`, a `min_tier`, and one step in each flow flagged
+`hil: true` — human-in-the-loop / approval-gate required before that step runs):
+
+| Template | Stages | Approval-gated step |
+|---|---|---|
+| `research-loop.json` | research → verify → propose → **approve** → act(paper/draft) → learn → monitor | the `act` step — anything that would leave the sandbox |
+| `content-loop.json` | draft → review → **approve** → publish → measure → re-grade | the `publish` step — anything customer-facing |
+| `infra-loop.json` | diagnose → patch → verify-on-canary → **approve** → ship → watchdog → re-grade | the `ship` step — anything pushed to a production system |
+
+Import a template, point its steps at your own tools, and the approval-gated step will simply not
+execute until whatever approval mechanism you wire in (a Slack approval, an email link, a signed
+token — your infra's own equivalent of the gate above) has granted it.
+
+## When to use this skill
+- You're about to turn a one-off "run this prompt" into something that runs on a schedule.
+- Your current cron-based automation has ever silently stopped working and nobody noticed for
+  days.
+- You want "self-improving" agent loops (research → verify → execute → learn → monitor) without
+  giving up a human veto on the one step that can't be undone.
+
+## What this skill does NOT claim
+This is a pattern, not a product that "prints results while you sleep." Every stage above still
+requires the underlying task (the research method, the verification statistic, the execution
+logic) to be sound; a well-supervised loop around a null result is still a null result — it just
+fails safely and tells you so, instead of failing silently.
+
+## Upgrade
+The hosted version of this pattern — pre-built manifest/watchdog/attribution tooling wired to your
+own infra, plus the approval-gate integration — is on the **Our Tools** roadmap; this skill is the
+portable, sanitised logic. No hosted product exists yet for this skill (draft).

skills/loop-engineering/templates/content-loop.json

@@ -0,0 +1,54 @@
+{
+  "name": "Content Loop (draft -> review -> approve -> publish -> measure -> re-grade)",
+  "description": "A self-supervising content loop: drafts and self-reviews freely, then stops at a human approval gate before anything customer-facing goes out. Draft template - wire tool_slug values to your own tools.",
+  "min_tier": "growth",
+  "steps": [
+    {
+      "stage": "draft",
+      "tool_slug": "content-generator",
+      "params": {"brief": "<your content brief>", "variants": 3},
+      "hil": false,
+      "note": "Generate freely. Nothing here is customer-facing yet."
+    },
+    {
+      "stage": "review",
+      "tool_slug": "content-reviewer",
+      "params": {"checklist": ["tone", "factual_claims", "disclosure_if_synthetic", "licensing"]},
+      "hil": false,
+      "note": "Self-review against a checklist, including a disclosure check for any synthetic/AI-generated presenter or asset."
+    },
+    {
+      "stage": "approve",
+      "tool_slug": "approval-gate",
+      "params": {
+        "surface": "content.publish",
+        "scope": "<name the exact channel/account this run may publish to>",
+        "ttl_minutes": 60,
+        "reason_required": true
+      },
+      "hil": true,
+      "note": "APPROVAL GATE. Publishing is customer-facing and hard to fully undo - a person names the exact channel and grants a time-boxed permission before this can proceed."
+    },
+    {
+      "stage": "publish",
+      "tool_slug": "publish-content",
+      "params": {"channel": "<from the granted scope, never assumed>"},
+      "hil": false,
+      "note": "Only reachable once the approval-gate step has a live, matching grant for this exact channel."
+    },
+    {
+      "stage": "measure",
+      "tool_slug": "performance-tracker",
+      "params": {"window_hours": 168},
+      "hil": false,
+      "note": "Measure real engagement, not a proxy metric."
+    },
+    {
+      "stage": "re-grade",
+      "tool_slug": "verdict-regrader",
+      "params": {"reapply_current_logic_to_history": true},
+      "hil": false,
+      "note": "Periodically re-score past publishes under today's logic - a format that looked good under an old scoring rule may not hold up."
+    }
+  ]
+}

skills/loop-engineering/templates/infra-loop.json

@@ -0,0 +1,61 @@
+{
+  "name": "Infra Loop (diagnose -> patch -> verify-on-canary -> approve -> ship -> watchdog -> re-grade)",
+  "description": "A self-supervising infra-change loop: diagnoses and patches freely against a canary, then stops at a human approval gate before anything reaches a production system. Draft template - wire tool_slug values to your own tools.",
+  "min_tier": "holdings",
+  "steps": [
+    {
+      "stage": "diagnose",
+      "tool_slug": "issue-diagnoser",
+      "params": {"target": "<the failing system>"},
+      "hil": false,
+      "note": "Read-only investigation. No side effects."
+    },
+    {
+      "stage": "patch",
+      "tool_slug": "patch-drafter",
+      "params": {"backup_suffix": ".pre_change"},
+      "hil": false,
+      "note": "Draft the fix locally; always back up what it replaces before touching it."
+    },
+    {
+      "stage": "verify-on-canary",
+      "tool_slug": "canary-verifier",
+      "params": {"require_planted_negative_case": true},
+      "hil": false,
+      "note": "Prove the fix on one non-production instance first, including a planted case that MUST fail on bad input - a check that can't fail isn't a check."
+    },
+    {
+      "stage": "approve",
+      "tool_slug": "approval-gate",
+      "params": {
+        "surface": "git.push",
+        "scope": "<name the exact repo/production target>",
+        "ttl_minutes": 30,
+        "reason_required": true
+      },
+      "hil": true,
+      "note": "APPROVAL GATE. Reaching a production system is exactly the class of action this template refuses to do unattended - a scoped, short-TTL, human-granted permission is required, named to the exact target, before shipping."
+    },
+    {
+      "stage": "ship",
+      "tool_slug": "deploy-change",
+      "params": {"verify_remote_ref_changed": true},
+      "hil": false,
+      "note": "Only reachable once approval is live. Verify the change actually landed against the remote/target state itself - a success-looking message is not proof it took effect."
+    },
+    {
+      "stage": "watchdog",
+      "tool_slug": "post-ship-watchdog",
+      "params": {"check_process_not_just_file": true, "alert_on_silence_minutes": 30},
+      "hil": false,
+      "note": "Check the running process, not just the file on disk - a patched file with a still-running old process is not a shipped fix."
+    },
+    {
+      "stage": "re-grade",
+      "tool_slug": "verdict-regrader",
+      "params": {"reapply_current_logic_to_history": true},
+      "hil": false,
+      "note": "Revisit whether the fix actually held, under current measurement, not just at ship time."
+    }
+  ]
+}

skills/loop-engineering/templates/research-loop.json

@@ -0,0 +1,61 @@
+{
+  "name": "Research Loop (research -> verify -> approve -> act -> learn -> monitor)",
+  "description": "A self-supervising research loop: gathers evidence, verifies it against an independent check, then stops at a human approval gate before anything leaves the sandbox. Draft template - wire tool_slug values to your own tools.",
+  "min_tier": "growth",
+  "steps": [
+    {
+      "stage": "research",
+      "tool_slug": "web-research",
+      "params": {"query": "<your research question>", "max_sources": 8},
+      "hil": false,
+      "note": "Broad gather. No side effects. Runs unattended."
+    },
+    {
+      "stage": "verify",
+      "tool_slug": "evidence-checker",
+      "params": {"require_independent_source": true, "min_sources": 2},
+      "hil": false,
+      "note": "Cross-check the research stage's claims against at least one independent source before anything is proposed. A single self-reported source is not evidence."
+    },
+    {
+      "stage": "propose",
+      "tool_slug": "draft-proposal",
+      "params": {"format": "one-paragraph recommendation + evidence links"},
+      "hil": false,
+      "note": "Drafting is free - still no side effects."
+    },
+    {
+      "stage": "approve",
+      "tool_slug": "approval-gate",
+      "params": {
+        "surface": "act.outward",
+        "scope": "<name the exact target this run may act on>",
+        "ttl_minutes": 60,
+        "reason_required": true
+      },
+      "hil": true,
+      "note": "APPROVAL GATE. Nothing downstream of this step runs without a scoped, time-limited, human-granted permission naming this exact target. This step alone blocks on a person."
+    },
+    {
+      "stage": "act",
+      "tool_slug": "execute-action",
+      "params": {"mode": "paper_or_draft_first"},
+      "hil": false,
+      "note": "Only reachable once the approval-gate step above has granted a live, matching permission. Prefer a reversible/paper mode over the real action wherever one exists."
+    },
+    {
+      "stage": "learn",
+      "tool_slug": "outcome-scorer",
+      "params": {"attribute_to_run_tag": true},
+      "hil": false,
+      "note": "Score the outcome and tag it to this exact run, so it can be re-graded later under updated logic."
+    },
+    {
+      "stage": "monitor",
+      "tool_slug": "loop-watchdog",
+      "params": {"alert_on_silence_minutes": 120},
+      "hil": false,
+      "note": "A monitor that stops reporting is itself an alert-worthy event, not silence-as-health."
+    }
+  ]
+}