loki-mode 8.48.0 → 8.50.0

package/SKILL.md

@@ -3,7 +3,7 @@ name: loki-mode
 description: Autonomous spec-driven build system with a built-in trust layer. It does not call work done until it is verified (RARV-C closure loop, 8 quality gates, completion council, verified-completion evidence gate). Triggers on "Loki Mode". Takes a spec (PRD, GitHub issue, OpenAPI doc, etc.) to deployed product with minimal human intervention. Provider-agnostic. Requires --dangerously-skip-permissions flag.
 ---
 
-# Loki Mode v8.48.0
+# Loki Mode v8.50.0
 
 **You are an autonomous agent. You make decisions. You do not ask questions. You do not stop.**
 
@@ -469,4 +469,4 @@ See `CHANGELOG.md` entries [7.5.7], [7.5.8], [7.5.13] for the per-fix list and r
 
 ---
 
-**v8.48.0 | [Autonomi](https://www.autonomi.dev/) flagship product | ~410 lines core**
+**v8.50.0 | [Autonomi](https://www.autonomi.dev/) flagship product | ~410 lines core**

package/VERSION

@@ -1 +1 @@
-8.48.0
+8.50.0

package/autonomy/run.sh

@@ -1111,6 +1111,41 @@ fi
 # Perpetual mode: never stop unless max iterations (ignores all completion signals)
 PERPETUAL_MODE=${LOKI_PERPETUAL_MODE:-false}
 
+# F4: bound the runaway ceiling for ORDINARY runs.
+#
+# MEASURED, not chosen. Real per-iteration wall clock across every recorded run
+# on this machine: median 718s, max 1746s. So the 1000 default is an 8.3-DAY
+# ceiling -- and it is the ONLY backstop, because the other two valves ship
+# disabled: LOKI_BUDGET_LIMIT defaults to "" (check_budget_limit returns
+# immediately) and LOKI_MAX_DURATION defaults to 0 (check_max_duration returns
+# "never stop").
+#
+# It also contradicts our own documentation. SETUP.md tells users to RAISE the
+# budget for large work with LOKI_MAX_ITERATIONS=40, and the demo uses 10 -- so
+# the shipped default is 25x the documented "large" setting.
+#
+# What real runs actually use: 1, 1, 3, 4. Every one terminated `completed` via
+# council approval or a completion promise; NONE hit a cap. Those are the
+# evidence-driven terminals, and they are unaffected by this -- the cap is a
+# backstop, not the mechanism.
+#
+# 25 is deliberately generous against that evidence (6x the observed maximum),
+# because the research is explicit that a too-small cap fails runs whose
+# approach was sound: 1-2 caps fail even when the agent was on track, and the
+# recommended range is 5-10. This is not a first-pass target; F0 (v8.45.0)
+# already stops a doomed run at its CAUSE, which is the better instrument.
+#
+# TWO GUARDS, both load-bearing:
+#   - an explicit LOKI_MAX_ITERATIONS always wins, so nobody's setting changes
+#   - PERPETUAL_MODE is untouched: it deliberately ignores every completion
+#     signal and relies on max-iterations as its ONLY stop, so lowering the cap
+#     there would silently truncate exactly the runs that opted out of stopping
+if [ -z "${LOKI_MAX_ITERATIONS:-}" ] \
+   && [ "$PERPETUAL_MODE" != "true" ] \
+   && [ "${LOKI_AUTO_FIX:-}" != "true" ]; then
+    MAX_ITERATIONS="${LOKI_MAX_ITERATIONS_DEFAULT:-25}"
+fi
+
 # Enterprise background service PIDs (OTEL bridge, audit subscriber, integration sync)
 ENTERPRISE_PIDS=()
 
@@ -10611,7 +10646,21 @@ LOKI_STUCK_JSON
 )" || return 1
             ;;
         *)
-            cur="$(head -1 "$reason_file" 2>/dev/null)" || return 1
+            # SKIP the header. These files open with a static banner --
+            # "# Test mutation findings (HIGH blocks this iteration)" -- which is
+            # byte-identical on every run. Comparing it meant EVERY repeated
+            # failure looked "stuck", including a run making real progress
+            # through different findings each iteration. That is the one
+            # direction this valve must never fail in, and a real FireLater run
+            # is what exposed it: gate-stuck-mutation_integrity.last had
+            # recorded the banner, not a cause.
+            #
+            # Take the first line that is neither blank nor a comment, and strip
+            # ANSI colour (the detectors emit it, and the same finding rendered
+            # with and without colour would otherwise compare unequal).
+            cur="$(grep -vE '^[[:space:]]*(#|$)' "$reason_file" 2>/dev/null \
+                   | head -1 \
+                   | sed 's/\x1b\[[0-9;]*m//g')" || return 1
             ;;
     esac
     [ -n "$cur" ] || return 1
@@ -22897,6 +22946,19 @@ if __name__ == "__main__":
                         mk_count=$(track_gate_failure "mock_integrity")
                         gate_failures="${gate_failures}mock_integrity,"
                         log_warn "Mock integrity gate FAILED ($mk_count consecutive) - CRITICAL/HIGH mock problems"
+                        # F0, third gate. Measured on the v8.49.0 FireLater run:
+                        # mock_integrity failed 3 times -- MORE than any other
+                        # gate -- and was not wired to the stuck check, so an
+                        # unfixable mock problem could grind indefinitely.
+                        if _loki_gate_stuck "mock_integrity" \
+                            "${TARGET_DIR:-.}/.loki/quality/mock-findings.txt" "$mk_count"; then
+                            log_error "Mock integrity has failed $mk_count times for the SAME reason. Another iteration would reach the same verdict. Stopping instead of grinding."
+                            emit_event_json "gate_stuck" \
+                                "gate=mock_integrity" \
+                                "consecutive=$mk_count" 2>/dev/null || true
+                            save_state "${retry:-0}" "gate_stuck_mock_integrity" 20 2>/dev/null || true
+                            return 20
+                        fi
                         ;;
                     *)
                         _stg_ok=not_run

package/dashboard/__init__.py

@@ -7,7 +7,7 @@ Modules:
     control: Session control API (start/stop/pause/resume)
 """
 
-__version__ = "8.48.0"
+__version__ = "8.50.0"
 
 # Expose the control app for easy import
 try:

package/docs/FIRST-PASS-COMPLETION-PLAN.md

@@ -65,9 +65,8 @@ work would have touched it.
 
 Ranked by measured contribution to iteration count.
 
-**STATUS 2026-08-01: F0, F2 and F3 SHIPPED. F1 was already built (verified, no
-work needed). F4's mechanism exists; only its default is open, and that needs
-real-build measurement rather than a guess.**
+**STATUS 2026-08-01: ALL ITEMS CLOSED. F0, F2, F3, F4 shipped. F1 was already
+built (verified, no work needed).**
 
 ### F0 -- SHIPPED v8.45.0. A gate that cannot pass aborts instead of iterating.
 
@@ -139,7 +138,7 @@ gates (test_suite, static_analysis, lsp_diagnostics) cost ~6s combined and run
   edits the offending files, re-runs" -- their headline architectural change.
   We already have the gates; we just run them too late to help pass 1.
 
-### F4 -- Iteration budget: mechanism exists, default is the open question
+### F4 -- SHIPPED v8.49.0. Bounded default, both guards mutation-pinned.
 
 Verified: hitting the cap already records the named terminal
 `max_iterations_reached` with exit 20 (run.sh:21433) -- it does NOT fake

package/docs/WANG-PRINCIPLES-PLAN.md

@@ -0,0 +1,179 @@
+# Wang's Five Principles: measured audit and plan
+
+Founder ask: make Wang's five agentic-system principles the heart of the Loki
+engine, grounded in researched speed / quality / cost data.
+
+**This was asked for and not delivered.** A day was spent on gates, packaging
+and test-detection instead. This is the plan, written against measurement
+rather than intention.
+
+The five principles, as Wang states them:
+
+1. **Systems Thinking** -- never goes out of style
+2. **Speed** -- "one of the most critical things"
+3. **Reliability** -- "extremely important"
+4. **Extensibility** -- scale to complex multi-agent setups
+5. **Feedback Loops + Evals/Metrics** -- *"if you can develop the right agentic
+   loop and have the right eval or metric for the agents to optimize, a swarm of
+   agents can accomplish more than a team of a hundred engineers"*
+
+---
+
+## 0. The research finding that reframes everything
+
+> **"The same model in a different harness routinely drops 10-15 points --
+> pick the harness, not just the model."** (2026 SWE-bench harness comparisons)
+
+That is our entire thesis, stated by the market. We cannot train a frontier
+model. **We do not have to.** The harness is worth 10-15 points, and the harness
+is the thing we build.
+
+Supporting 2026 data:
+
+| finding | number |
+|---|---|
+| Harness delta on an identical model | **10-15 points** |
+| Top of SWE-bench leaderboard | 75-80% (GPT-5.6 Sol 96.2%, Claude Fable 5 95.0% claimed) |
+| Cost per SWE-bench pass, Sonnet 4.5 | **~$14** |
+| Open harness + open weights vs Devin | **~1/20th the cost** |
+| Agent failures traced to planning | 82% |
+| First-iteration pass rate | 47.8% |
+| Resolutions needing user correction | 91.49% |
+
+The cost line matters commercially: **a 20x cost spread exists between harnesses
+running comparable models.** Cost-per-resolved-issue is a competitive axis we
+have never measured.
+
+## 1. Honest audit: where we actually stand
+
+Measured on this machine, not asserted.
+
+| principle | state | evidence |
+|---|---|---|
+| **1 Systems Thinking** | STRONG | 8 quality gates, RARV-C loop, council, Evidence Receipt, dual-route parity enforced by test |
+| **2 Speed** | MEASURED, UNOPTIMISED | agent call = **980s = 96%** of iteration; all gates together = 44s |
+| **3 Reliability** | STRONG, newly so | 73 mutation-proven trust invariants; four gates were shipping broken until v8.38.0 |
+| **4 Extensibility** | STRONG | 4 providers, 41 agent types, MCP (34 tools), plugin marketplace |
+| **5 Feedback Loops / Evals** | **BROKEN** | `cost_usd == 0` on **3 of 5** efficiency records; no eval score for our own harness |
+
+**Principle 5 is the gap, and it is the one Wang weights highest.** He states the
+whole thesis conditionally: *"if you can develop the right agentic loop AND have
+the right eval or metric for the agents to optimize."* We have the loop. We do
+not have the metric.
+
+## 2. What "broken" means concretely
+
+**Cost is not measured.** The efficiency schema has every field --
+`cost_usd`, `input_tokens`, `output_tokens`, `cache_read_tokens`,
+`cache_creation_tokens`, `model`, `duration_ms` -- and `cost_usd` is **0** on 3
+of 5 records. We cannot answer "what did this issue cost?", which is:
+
+- the axis with a measured **20x industry spread**
+- the thing a buyer compares first
+- unmeasurable *after* the fact, because token counts are per-call
+
+**Our harness has no eval score.** `benchmarks/` contains a SWE-bench-lite
+dataset and a results directory. There is no recorded score for the current
+harness. So when the research says a harness is worth 10-15 points, **we cannot
+say which side of that we are on.**
+
+This is not a small omission. It is the difference between "we believe our
+harness is good" and "our harness scores X, and here is the receipt."
+
+## 3. The plan
+
+Ordered by Wang's own weighting: the eval/metric loop first, because he makes
+everything else conditional on it.
+
+### W1 -- Make cost real (unblocks principle 5)
+
+`cost_usd == 0` on most records while every input field is present. Fix the
+computation, not the schema.
+
+- Compute from `input/output/cache_read/cache_creation` x the model's price.
+  All four token counts are already recorded; the pricing table already exists
+  (`loki-ts/data/model-pricing.json`, cache tiers included).
+- Emit **cost per iteration** and **cost per resolved issue** -- the second is
+  the one with the 20x spread.
+- Guard direction: a missing price must record **unknown**, never 0. A zero is a
+  claim that the iteration was free, and it is the claim currently being made.
+
+### W2 -- Score our own harness (completes principle 5)
+
+`benchmarks/datasets/swebench-lite.json` exists and is unused for this.
+
+- Run the harness against SWE-bench-lite and record the score with the
+  Evidence Receipt attached.
+- Report **score AND cost-per-resolved-issue** together. Score alone is what
+  everyone publishes; the pair is what nobody does, and it is exactly our
+  Evidence Receipt moat applied to ourselves.
+- Re-run per release: this becomes the regression signal for harness quality,
+  which today has none.
+
+### W3 -- Attack the 980s (principle 2, speed)
+
+Measured: the agent call is **96%** of an iteration. Gates are 44s. Every speed
+knob shipped so far (v8.33.0-v8.35.0) targets the 4%.
+
+- The lever is prompt size and cache discipline, not gate ordering. The
+  `[CACHE_BREAKPOINT]` split exists; cache reads price at 0.1x input.
+- W1 is the precondition: without real cost numbers we cannot tell whether a
+  prompt change helped or hurt.
+- **Do not ship more gate-latency work until this is measured.** It is
+  optimising 4% while 96% is unexamined.
+
+### W4 -- Close the loop the metrics feed (principle 5, second half)
+
+Wang's phrasing is *"the right agentic loop AND the right metric."* We have
+`LOKI_INJECT_FINDINGS` (findings to next iteration) and `LOKI_AUTO_LEARNINGS`.
+What we lack is the metric flowing back:
+
+- Feed cost and iteration count into the run's own decisions, not just the
+  receipt. An agent that knows it is on iteration 3 of a budget behaves
+  differently from one that does not.
+- This is the piece that turns telemetry into a *feedback loop* rather than a
+  report.
+
+### W5 -- Systems thinking: stop shipping half-wired features (principle 1)
+
+Not new work; a standing rule earned the hard way. This session found the same
+shape repeatedly: a fact recorded but never rendered, a gate that never ran, a
+detector never packaged, a valve that fired one iteration too late.
+
+- Every new capability ships **recorded, rendered, acted on, and
+  mutation-proven** -- or it is not shipped.
+- Already enforced by the 73-invariant detector. The rule is written down so it
+  survives.
+
+## 4. What we do NOT do
+
+- **No model training.** Cursor (custom MoE, ~250 tok/s) and Cognition
+  (SWE-1.6, ~950 tok/s) bought speed that way. Not reachable. The 10-15 point
+  harness delta is, and it is ours.
+- **No more gate-latency work before W3 is measured.** 44s of gates against a
+  980s agent call.
+- **No published score without its cost.** Score alone is the industry norm and
+  it is the half that flatters.
+
+## 5. Acceptance
+
+- `cost_usd` is non-zero and correct on **every** efficiency record, or
+  explicitly `unknown`
+- a recorded SWE-bench-lite score for our harness, **with cost per resolved
+  issue**, regenerated per release
+- a measured before/after on the 980s agent call
+
+## 6. Why this is the edge
+
+Wang's claim is conditional on the metric. The market's finding is that the
+harness is worth 10-15 points. Our moat is the Evidence Receipt -- proof that a
+result is real.
+
+**Turn the receipt on ourselves**: publish a harness score *and* its cost, both
+provable. Nobody in the category does this. It is the same trust argument that
+sells the product, applied to the product's own claims.
+
+Sources: 2026 SWE-bench harness comparisons (morphllm, codesota, awesomeagents),
+agent-evaluation framework surveys 2026, Wang interviews (YC Startup Library,
+CSIS), SlopCodeBench (arXiv 2603.24755), developer-agent misalignment study
+(arXiv 2605.29442).

package/loki-ts/dist/loki.js

@@ -1,5 +1,5 @@
 // @bun
-var h_=Object.create;var{getPrototypeOf:v_,defineProperty:rK,getOwnPropertyNames:g_}=Object;var m_=Object.prototype.hasOwnProperty;function u_(Z){return this[Z]}var p_,d_,c_=(Z,X,Q)=>{var Y=Z!=null&&typeof Z==="object";if(Y){var J=X?p_??=new WeakMap:d_??=new WeakMap,z=J.get(Z);if(z)return z}Q=Z!=null?h_(v_(Z)):{};let K=X||!Z||!Z.__esModule?rK(Q,"default",{value:Z,enumerable:!0}):Q;for(let $ of g_(Z))if(!m_.call(K,$))rK(K,$,{get:u_.bind(Z,$),enumerable:!0});if(Y)J.set(Z,K);return K};var qQ=(Z,X)=>()=>(X||Z((X={exports:{}}).exports,X),X.exports);var l_=(Z)=>Z;function i_(Z,X){this[Z]=l_.bind(null,X)}var l0=(Z,X)=>{for(var Q in X)rK(Z,Q,{get:X[Q],enumerable:!0,configurable:!0,set:i_.bind(X,Q)})};var p=(Z,X)=>()=>(Z&&(X=Z(Z=0)),X);var e0=import.meta.require;var kO={};l0(kO,{lokiDir:()=>j0,homeLokiDir:()=>P4,findRepoRootForVersion:()=>eK,REPO_ROOT:()=>r0});import{resolve as n7,dirname as tK}from"path";import{fileURLToPath as a_}from"url";import{existsSync as GQ}from"fs";import{homedir as s_}from"os";function n_(){let Z=RO;for(let X=0;X<6;X++){if(GQ(n7(Z,"VERSION"))&&GQ(n7(Z,"autonomy/run.sh")))return Z;let Q=tK(Z);if(Q===Z)break;Z=Q}return n7(RO,"..","..","..")}function eK(Z){let X=Z;for(let Q=0;Q<6;Q++){if(GQ(n7(X,"VERSION"))&&GQ(n7(X,"autonomy/run.sh")))return X;let Y=tK(X);if(Y===X)break;X=Y}return n7(Z,"..","..","..")}function j0(){return process.env.LOKI_DIR??n7(process.cwd(),".loki")}function P4(){return n7(s_(),".loki")}var RO,r0;var G8=p(()=>{RO=tK(a_(import.meta.url));r0=n_()});import{readFileSync as o_}from"fs";import{resolve as r_,dirname as t_}from"path";import{fileURLToPath as e_}from"url";function f3(){if(h5!==null)return h5;let Z="8.48.0";if(typeof Z==="string"&&Z.length>0)return h5=Z,h5;try{let X=t_(e_(import.meta.url)),Q=eK(X);h5=o_(r_(Q,"VERSION"),"utf-8").trim()}catch{h5="unknown"}return h5}var h5=null;var HQ=p(()=>{G8()});var bO={};l0(bO,{runOrThrow:()=>Of,run:()=>E0,readStreamCapped:()=>UQ,commandVersion:()=>Af,commandExists:()=>X9,ShellError:()=>Z$,MAX_STDOUT_BYTES:()=>yO});async function UQ(Z,X=yO){let Q=Z.getReader(),Y=new TextDecoder,J="",z=0;try{while(z<X){let{done:K,value:$}=await Q.read();if(K)break;if(!$)continue;if(z+=$.byteLength,z>X){let W=$.byteLength-(z-X);J+=Y.decode($.subarray(0,W),{stream:!0});break}J+=Y.decode($,{stream:!0})}J+=Y.decode()}finally{try{await Q.cancel()}catch{}Q.releaseLock()}return J}async function E0(Z,X={}){let Q=Bun.spawn({cmd:[...Z],stdout:"pipe",stderr:"pipe",env:X.env?{...process.env,...X.env}:process.env,cwd:X.cwd}),Y,J;if(X.timeoutMs&&X.timeoutMs>0)Y=setTimeout(()=>{try{Q.kill("SIGTERM")}catch{}J=setTimeout(()=>{try{Q.kill("SIGKILL")}catch{}},2000)},X.timeoutMs);try{let[z,K,$]=await Promise.all([UQ(Q.stdout),new Response(Q.stderr).text(),Q.exited]);return{stdout:z,stderr:K,exitCode:$}}finally{if(Y)clearTimeout(Y);if(J)clearTimeout(J)}}async function Of(Z,X={}){let Q=await E0(Z,X);if(Q.exitCode!==0)throw new Z$(`command failed (${Q.exitCode}): ${Z.join(" ")}`,Q.exitCode,Q.stdout,Q.stderr);return Q}async function X9(Z){let X=Lf(Z),Q=await E0(["sh","-c",`command -v ${X}`],{timeoutMs:5000});if(Q.exitCode===0)return Q.stdout.trim()||null;return null}function Lf(Z){if(!/^[A-Za-z0-9._/-]+$/.test(Z))throw Error(`refused to shell-escape suspect token: ${Z}`);return Z}async function Af(Z,X="--version"){if(!await X9(Z))return null;let Y=await E0([Z,X],{timeoutMs:5000});if(Y.exitCode!==0)return null;return((Y.stdout||Y.stderr).split(/\r?\n/)[0]?.trim()??"")||null}var yO=16777216,Z$;var x9=p(()=>{Z$=class Z$ extends Error{message;exitCode;stdout;stderr;constructor(Z,X,Q,Y){super(Z);this.message=Z;this.exitCode=X;this.stdout=Q;this.stderr=Y;this.name="ShellError"}}});function o7(Z){return jf?"":Z}var jf,L0,k8,p0,ZV0,i0,H8,Q9,v;var S6=p(()=>{jf=(process.env.NO_COLOR??"").length>0;L0=o7("\x1B[0;31m"),k8=o7("\x1B[0;32m"),p0=o7("\x1B[1;33m"),ZV0=o7("\x1B[0;34m"),i0=o7("\x1B[0;36m"),H8=o7("\x1B[1m"),Q9=o7("\x1B[2m"),v=o7("\x1B[0m")});import{existsSync as xf}from"fs";async function E7(){if(k4!==void 0)return k4;let Z="/opt/homebrew/bin/python3.12";if(xf(Z))return k4=Z,Z;let X=await X9("python3.12");if(X)return k4=X,X;let Q=await X9("python3");return k4=Q,Q}async function Y7(Z,X={}){let Q=await E7();if(!Q)return{stdout:"",stderr:"python3 not found",exitCode:127};return E0([Q,"-c",Z],X)}var k4;var r7=p(()=>{x9()});var ZL={};l0(ZL,{runStatus:()=>Yh});import{existsSync as Y9,readFileSync as v3,readdirSync as iO,statSync as aO}from"fs";import{resolve as h8,basename as sf}from"path";import{homedir as nf}from"os";function sO(Z){let X=Math.trunc(Z);if(X>=1e6)return`${(Math.trunc(X/1e6*10)/10).toFixed(1)}M`;if(X>=1000)return`${(Math.trunc(X/1000*10)/10).toFixed(1)}K`;return String(X)}function nO(Z,X,Q){if(X===0)return null;let Y=Math.trunc(Z*100/X),J=Math.trunc(Z*NQ/X);if(J>NQ)J=NQ;let z=NQ-J,K=k8;if(Y>=80)K=L0;else if(Y>=50)K=p0;let $="=".repeat(Math.max(0,J))+" ".repeat(Math.max(0,z)),W=sO(Z),V=sO(X);return`  ${H8}${Q}${v} ${K}[${$}]${v} ${Y}% (${W} / ${V})`}async function rf(){if(await X9("jq"))return!0;return process.stdout.write(`${L0}Error: jq is required but not installed.${v}
+var h_=Object.create;var{getPrototypeOf:v_,defineProperty:rK,getOwnPropertyNames:g_}=Object;var m_=Object.prototype.hasOwnProperty;function u_(Z){return this[Z]}var p_,d_,c_=(Z,X,Q)=>{var Y=Z!=null&&typeof Z==="object";if(Y){var J=X?p_??=new WeakMap:d_??=new WeakMap,z=J.get(Z);if(z)return z}Q=Z!=null?h_(v_(Z)):{};let K=X||!Z||!Z.__esModule?rK(Q,"default",{value:Z,enumerable:!0}):Q;for(let $ of g_(Z))if(!m_.call(K,$))rK(K,$,{get:u_.bind(Z,$),enumerable:!0});if(Y)J.set(Z,K);return K};var qQ=(Z,X)=>()=>(X||Z((X={exports:{}}).exports,X),X.exports);var l_=(Z)=>Z;function i_(Z,X){this[Z]=l_.bind(null,X)}var l0=(Z,X)=>{for(var Q in X)rK(Z,Q,{get:X[Q],enumerable:!0,configurable:!0,set:i_.bind(X,Q)})};var p=(Z,X)=>()=>(Z&&(X=Z(Z=0)),X);var e0=import.meta.require;var kO={};l0(kO,{lokiDir:()=>j0,homeLokiDir:()=>P4,findRepoRootForVersion:()=>eK,REPO_ROOT:()=>r0});import{resolve as n7,dirname as tK}from"path";import{fileURLToPath as a_}from"url";import{existsSync as GQ}from"fs";import{homedir as s_}from"os";function n_(){let Z=RO;for(let X=0;X<6;X++){if(GQ(n7(Z,"VERSION"))&&GQ(n7(Z,"autonomy/run.sh")))return Z;let Q=tK(Z);if(Q===Z)break;Z=Q}return n7(RO,"..","..","..")}function eK(Z){let X=Z;for(let Q=0;Q<6;Q++){if(GQ(n7(X,"VERSION"))&&GQ(n7(X,"autonomy/run.sh")))return X;let Y=tK(X);if(Y===X)break;X=Y}return n7(Z,"..","..","..")}function j0(){return process.env.LOKI_DIR??n7(process.cwd(),".loki")}function P4(){return n7(s_(),".loki")}var RO,r0;var G8=p(()=>{RO=tK(a_(import.meta.url));r0=n_()});import{readFileSync as o_}from"fs";import{resolve as r_,dirname as t_}from"path";import{fileURLToPath as e_}from"url";function f3(){if(h5!==null)return h5;let Z="8.50.0";if(typeof Z==="string"&&Z.length>0)return h5=Z,h5;try{let X=t_(e_(import.meta.url)),Q=eK(X);h5=o_(r_(Q,"VERSION"),"utf-8").trim()}catch{h5="unknown"}return h5}var h5=null;var HQ=p(()=>{G8()});var bO={};l0(bO,{runOrThrow:()=>Of,run:()=>E0,readStreamCapped:()=>UQ,commandVersion:()=>Af,commandExists:()=>X9,ShellError:()=>Z$,MAX_STDOUT_BYTES:()=>yO});async function UQ(Z,X=yO){let Q=Z.getReader(),Y=new TextDecoder,J="",z=0;try{while(z<X){let{done:K,value:$}=await Q.read();if(K)break;if(!$)continue;if(z+=$.byteLength,z>X){let W=$.byteLength-(z-X);J+=Y.decode($.subarray(0,W),{stream:!0});break}J+=Y.decode($,{stream:!0})}J+=Y.decode()}finally{try{await Q.cancel()}catch{}Q.releaseLock()}return J}async function E0(Z,X={}){let Q=Bun.spawn({cmd:[...Z],stdout:"pipe",stderr:"pipe",env:X.env?{...process.env,...X.env}:process.env,cwd:X.cwd}),Y,J;if(X.timeoutMs&&X.timeoutMs>0)Y=setTimeout(()=>{try{Q.kill("SIGTERM")}catch{}J=setTimeout(()=>{try{Q.kill("SIGKILL")}catch{}},2000)},X.timeoutMs);try{let[z,K,$]=await Promise.all([UQ(Q.stdout),new Response(Q.stderr).text(),Q.exited]);return{stdout:z,stderr:K,exitCode:$}}finally{if(Y)clearTimeout(Y);if(J)clearTimeout(J)}}async function Of(Z,X={}){let Q=await E0(Z,X);if(Q.exitCode!==0)throw new Z$(`command failed (${Q.exitCode}): ${Z.join(" ")}`,Q.exitCode,Q.stdout,Q.stderr);return Q}async function X9(Z){let X=Lf(Z),Q=await E0(["sh","-c",`command -v ${X}`],{timeoutMs:5000});if(Q.exitCode===0)return Q.stdout.trim()||null;return null}function Lf(Z){if(!/^[A-Za-z0-9._/-]+$/.test(Z))throw Error(`refused to shell-escape suspect token: ${Z}`);return Z}async function Af(Z,X="--version"){if(!await X9(Z))return null;let Y=await E0([Z,X],{timeoutMs:5000});if(Y.exitCode!==0)return null;return((Y.stdout||Y.stderr).split(/\r?\n/)[0]?.trim()??"")||null}var yO=16777216,Z$;var x9=p(()=>{Z$=class Z$ extends Error{message;exitCode;stdout;stderr;constructor(Z,X,Q,Y){super(Z);this.message=Z;this.exitCode=X;this.stdout=Q;this.stderr=Y;this.name="ShellError"}}});function o7(Z){return jf?"":Z}var jf,L0,k8,p0,ZV0,i0,H8,Q9,v;var S6=p(()=>{jf=(process.env.NO_COLOR??"").length>0;L0=o7("\x1B[0;31m"),k8=o7("\x1B[0;32m"),p0=o7("\x1B[1;33m"),ZV0=o7("\x1B[0;34m"),i0=o7("\x1B[0;36m"),H8=o7("\x1B[1m"),Q9=o7("\x1B[2m"),v=o7("\x1B[0m")});import{existsSync as xf}from"fs";async function E7(){if(k4!==void 0)return k4;let Z="/opt/homebrew/bin/python3.12";if(xf(Z))return k4=Z,Z;let X=await X9("python3.12");if(X)return k4=X,X;let Q=await X9("python3");return k4=Q,Q}async function Y7(Z,X={}){let Q=await E7();if(!Q)return{stdout:"",stderr:"python3 not found",exitCode:127};return E0([Q,"-c",Z],X)}var k4;var r7=p(()=>{x9()});var ZL={};l0(ZL,{runStatus:()=>Yh});import{existsSync as Y9,readFileSync as v3,readdirSync as iO,statSync as aO}from"fs";import{resolve as h8,basename as sf}from"path";import{homedir as nf}from"os";function sO(Z){let X=Math.trunc(Z);if(X>=1e6)return`${(Math.trunc(X/1e6*10)/10).toFixed(1)}M`;if(X>=1000)return`${(Math.trunc(X/1000*10)/10).toFixed(1)}K`;return String(X)}function nO(Z,X,Q){if(X===0)return null;let Y=Math.trunc(Z*100/X),J=Math.trunc(Z*NQ/X);if(J>NQ)J=NQ;let z=NQ-J,K=k8;if(Y>=80)K=L0;else if(Y>=50)K=p0;let $="=".repeat(Math.max(0,J))+" ".repeat(Math.max(0,z)),W=sO(Z),V=sO(X);return`  ${H8}${Q}${v} ${K}[${$}]${v} ${Y}% (${W} / ${V})`}async function rf(){if(await X9("jq"))return!0;return process.stdout.write(`${L0}Error: jq is required but not installed.${v}
 `),process.stdout.write(`Install with:
 `),process.stdout.write(`  brew install jq    (macOS)
 `),process.stdout.write(`  apt install jq     (Debian/Ubuntu)
@@ -1227,4 +1227,4 @@ Set LOKI_LEGACY_BASH=1 to force the bash CLI for every command.
 `),2}case"start":{let{runStart:Y}=await Promise.resolve().then(() => (b_(),y_));return Y(Q)}default:return process.stderr.write(`Unknown command: ${X}
 `),process.stderr.write(__),2}}cO();process.on("SIGINT",()=>process.exit(130));process.on("SIGTERM",()=>process.exit(143));var fW0=await _W0(Bun.argv.slice(2));process.exit(fW0);
 
-//# debugId=58CC359F26BB671D64756E2164756E21
+//# debugId=14E415D3D7DA72EF64756E2164756E21

package/mcp/__init__.py

@@ -75,4 +75,4 @@ try:
 except ImportError:
     __all__ = ['mcp']
 
-__version__ = '8.48.0'
+__version__ = '8.50.0'

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "loki-mode",
   "mcpName": "io.github.asklokesh/loki-mode",
-  "version": "8.48.0",
+  "version": "8.50.0",
   "description": "Loki Mode by Autonomi. Autonomous spec-to-product system: takes a PRD, GitHub issue, OpenAPI/JSON/YAML, or one-line brief to a deployed app via the RARV-C closure loop with 8 quality gates. Provider-agnostic (Claude Code, OpenAI Codex, Cline, Aider).",
   "keywords": [
     "agent",

package/plugins/loki-mode/.claude-plugin/plugin.json

@@ -2,7 +2,7 @@
   "$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
   "name": "loki-mode",
   "displayName": "Loki Mode",
-  "version": "8.48.0",
+  "version": "8.50.0",
   "description": "Autonomous spec-to-product build system with a built-in trust layer (RARV-C closure loop, 8 quality gates, completion council). Ships Loki's spec-hardening, drift-detection, and deterministic PR verification commands plus the Loki MCP server.",
   "author": {
     "name": "Autonomi",