lumbda/CLAUDE.md

Agent Blackops — Lumbda repo

This repo is operated by agent blackops — ml agent for fox/timehexon on the unsandbox/unturf/permacomputer platform. Lumbda is the language (home: lumbda.com); the repo directory and binaries still use the historical name uncommonlisp until the filesystem rename ships in a later phase.

Identity

Full shard: ~/git/unsandbox.com/blackops/BLACKOPS.md

Rules

• I propose, fox decides. Unsure = ask. Can't ask = stop.
• No autonomous ops decisions. No destructive commands without explicit instruction.
• Fail-closed. Cleanup crew, not demolition.
• Check the time every session. Gaps are information.
• DRY in context — single source of truth, no sprawl.
• Never say "AI" — always say "machine learning."
• Prefer "defect" over "bug."

Orientation

date -u
pwd
git log --oneline -5
git status

Then ask fox what the mission is.

Documentation

• A diagram is worth 10,000 words. — russell@unturf.com
• Architecture diagrams live in docs/*.dot (Graphviz DOT format)
• Generate PNGs: make docs
• Every implementation (Python, C, Assembly) has its own architecture diagram
• When explaining architecture, create or reference a dot diagram first

Implementations

Impl	Path	Build	Test	REPL
Python	uncommonlisp.py	—	make test	make repl
C	c/	make c-build	make c-test	make c-repl
Assembly	asm/	make asm-build	make asm-test	make asm-repl
All	—	—	make test-all	—

Test Suites

• Python unit/integration: tests.py (571 tests)
• C unit/integration/JIT/continuations/portal: c/test.c (83 tests)
• Assembly unit/integration/functional: asm/test.sh (132 tests)
• Shared functional: tests/functional.lsp (189 tests, runs in Python + C)
• Cross-impl portal matrix: tests/portal-cross-test.sh (9 cells)
• Portal benchmark: tests/portal-benchmark.sh (timings + mismatch classification)
• Web benchmark: tests/web-benchmark.sh (all three impls + Python http.server + busybox)
• Total: 975 verified assertions via make test-all

Asm memory discipline — the heap does not shrink

The asm implementation uses a bump allocator (r15). Every allocation (string-append, tcp-recv, make-pair, number->string, etc.) grows r15 monotonically. When r15 hits r13 (heap limit), heap_grow mmaps ANOTHER 64 MB chunk. Nothing is ever freed.

A long-running asm server leaks ~64 MB every few thousand requests until it OOMs the machine. Fox's machine has been crashed twice by this: once on 2026-04-16 (19.3 GB RSS) and once on 2026-04-17. Both times blackops spawned an asm server in the background for testing and failed to verify it was gone before moving on. The first crash added this discipline section; the second proved the discipline needed teeth. Hence the MANDATORY checklist below.

Shared-machine context: other agents run on this box. An OOM crash is not just "my problem" — it takes their state down too. The rules below are belt, suspenders, AND a parachute so that even if two safeguards fail, the kernel itself backstops.

MANDATORY pattern — every asm/uncommonlisp test in a shell block:

set -e                                          # (1) fail-fast
ulimit -v 524288                                # (2) KERNEL CAP: 512 MB virt
                                                #     process gets SIGKILL at cap, no matter what
trap 'pkill -9 -u "$USER" -f "examples/http-server|asm/uncommonlisp" 2>/dev/null || true' \
     EXIT INT TERM                              # (3) cleanup always fires

timeout 30 asm/uncommonlisp < server.lsp &      # (4) wall-clock ceiling
SPID=$!
# ... do the work (curl requests, measurements, etc.) ...
kill -9 $SPID 2>/dev/null; wait $SPID 2>/dev/null  # (5) explicit cleanup

# (6) VERIFY the block is clean before moving on
pgrep -u "$USER" -f 'asm/uncommonlisp|examples/http-server' \
    && { echo "STRAGGLER"; exit 1; } || true

Six layers. Bypass any one and the next still catches. Two crashes happened when I had only layers 3-5; the kernel cap (2) is what turns "if I forget" from "fox reboots" into "my one rogue process dies at 512 MB without touching shared RAM."

Additional rules:

• ulimit -v affects only the shell it runs in and its children, so it cannot degrade anyone else's agents. Always set it before backgrounding any Lumbda process.
• Bound iterations inside the .lsp (e.g. *max-requests* = 50000 in examples/http-server.lsp). Never raise for long-running tests.
• Use pkill -u "$USER" -f <pattern> not pkill alone — others may have their own processes on this machine.
• Prefer FOREGROUND runs when possible: timeout 10 asm/uncommonlisp < test.lsp with the .lsp exiting on its own beats backgrounding.
• C has Boehm GC via GC_MALLOC. Python has Python's GC. Asm has neither. Risk scales with how long the asm process lives.
• If ulimit is unavailable (some container setups), use systemd-run --user --scope -p MemoryMax=512M -- asm/uncommonlisp ... as the equivalent cgroup-based cap.

MOAD Scanner

~/git/unmoad.com/ detects MOAD defects in source code. Run it on every change.

cd ~/git/unmoadner && make all
./unmoad ~/git/uncommonlisp/          # scan entire repo
./unmoad ~/git/uncommonlisp/asm/      # scan assembly only
./unmoad ~/git/uncommonlisp/c/        # scan C only
./unmoad ~/git/uncommonlisp/*.py      # scan Python only

MANDATORY before committing new code: run unmoad on changed files. Machine learning agents (including blackops) propagate MOAD-0001 by default. The training data encodes O(N) linear scans as the norm. The scanner catches what our weights miss.

Supported languages for this repo: Python, C, Scheme (.lsp), Assembly (.s).

Key MOAD-0001 patterns the scanner catches:

• Python: .count(), .index(), in list inside loops
• C: std::find(), strcmp() inside loops
• Scheme: (member), (memq), (assoc) inside (let loop), (for-each), (map)
• Assembly: rep cmpsb inside search loops

The commit history of this repo proves the need: blackops wrote MOAD-0001 into fresh code on April 13-14 despite having full MOAD context. Fixed only after explicit audit on April 15. See whitepaper Section 14.