Skip to content

Pixel-CLI/pixel / .agents/skills/docker-setup-smoke

docker-setup-smoke

Verified

Replay Pixel's new-user setup in Docker without an LLM — install through a release archive, install.sh, the Homebrew tap, or sources from main, a PR head or a pinned commit; then personal Claude/Codex/pi settings kept, idempotent reinstall, first pixel audit, doctor and --fix, project install, disabled classify and uninstall; with --agents, Claude Code, Codex and pi sessions against a scripted fake model show whether each receives the Pixel prompt, runs pixel and routes grep. Use for Linux setup smoke checks, not macOS behavior or real model inference.

FileKindSize
SKILL.mdskill md10.7 KB
scripts/Dockerfileother1.8 KB
scripts/agents.shscript7.6 KB
scripts/bootstrap.shscript1.0 KB
scripts/brew.shscript2.6 KB
scripts/checks.shscript11.8 KB
scripts/fake-llm.pyscript8.8 KB
scripts/installer.shscript934 B
scripts/run.shscript5.6 KB
scripts/source.shscript1.0 KB
scripts/test-runner.pyscript11.1 KB

scripts/checks.sh

View on GitHub
#!/bin/sh
# SPDX-FileCopyrightText: The Pixel contributors
# SPDX-License-Identifier: MIT

set -eu
# The bootstrap names the binary's directory and whether pixel owns it.
# shellcheck source=/dev/null
. /etc/pixel-smoke.env
export PATH="$PIXEL_BIN_DIR:$PATH"
export PIXEL_METRICS=0
test "$(command -v pixel)" = "$PIXEL_BIN_DIR/pixel"
report_ok() {
    python3 - "$1" <<'PY'
import json
import sys
with open(sys.argv[1]) as report:
    data = json.load(report)
assert data["ok"] is True, data
PY
}
# Every red check must say how to repair it.
reds_have_fixes() {
    python3 - "$1" <<'PY'
import sys
lines = open(sys.argv[1]).read().splitlines()
reds = [i for i, line in enumerate(lines) if line.lstrip().startswith("[red]")]
assert reds, "expected red checks in a repository that was never prepared"
for i in reds:
    assert i + 1 < len(lines) and lines[i + 1].lstrip().startswith("fix: "), lines[i]
PY
}
managed_state() {
    find "$HOME/.claude" "$HOME/.codex" "$HOME/.pi" "$HOME/.local/share/pixel" "$HOME/.bashrc" \
        -type f ! -name '*.pixel-bak.*' -exec sha256sum {} + 2>/dev/null | sort
}
# Personal settings a user had before Pixel: install keeps them, uninstall restores them.
mkdir -p "$HOME/.claude" "$HOME/.codex" "$HOME/.pi/agent"
cat > "$HOME/.claude/settings.json" <<'JSON'
{
  "model": "opus",
  "permissions": {"allow": ["Bash(ls:*)"]},
  "hooks": {
    "PreToolUse": [{"matcher": "Edit|Write", "hooks": [{"type": "command", "command": "/usr/local/bin/my-edit-guard"}]}],
    "SessionStart": [{"hooks": [{"type": "command", "command": "echo my-session-hook"}]}]
  }
}
JSON
printf '# my global claude rules\n' > "$HOME/.claude/CLAUDE.md"
printf 'model = "gpt-5"\n\n[profiles.work]\nmodel = "o3"\n' > "$HOME/.codex/config.toml"
printf '{"hooks": {"SessionStart": [{"hooks": [{"type": "command", "command": "echo my-codex-hook"}]}]}}\n' \
    > "$HOME/.codex/hooks.json"
printf '# my codex rules\n' > "$HOME/.codex/AGENTS.md"
printf '{"defaultProvider": "anthropic"}\n' > "$HOME/.pi/agent/settings.json"
printf '# my pi rules\n' > "$HOME/.pi/agent/AGENTS.md"
printf '# my pi system append\n' > "$HOME/.pi/agent/APPEND_SYSTEM.md"
mkdir /evidence/personal-before
(cd "$HOME" && cp -R --parents .claude .codex .pi /evidence/personal-before/)
(cd /evidence/personal-before && find . -type f | sort) > /evidence/personal-files.txt
personal_kept() {
    python3 - /evidence/personal-before "$1" <<'PY'
import json
import pathlib
import sys
before, phase = pathlib.Path(sys.argv[1]), sys.argv[2]
home = pathlib.Path.home()
def load(root, rel):
    return json.loads((root / rel).read_text())
for rel in (".claude/settings.json", ".codex/hooks.json"):
    old, new = load(before, rel), load(home, rel)
    if phase == "uninstalled":
        assert new == old, (rel, new)
        continue
    for key, value in old.items():
        if key != "hooks":
            assert new.get(key) == value, (rel, key, new.get(key))
    for event, groups in old["hooks"].items():
        for group in groups:
            assert group in new["hooks"].get(event, []), (rel, event, group)
for rel in (".claude/CLAUDE.md", ".codex/AGENTS.md", ".pi/agent/AGENTS.md", ".pi/agent/settings.json"):
    assert (home / rel).read_bytes() == (before / rel).read_bytes(), rel
for rel in (".codex/config.toml", ".pi/agent/APPEND_SYSTEM.md"):
    old, new = (before / rel).read_text(), (home / rel).read_text()
    if phase == "uninstalled":
        assert new == old, (rel, new)
    else:
        assert all(line in new.splitlines() for line in old.splitlines()), (rel, new)
PY
}
mkdir "$HOME/project"
cd "$HOME/project"
git init -q
git config user.email smoke@example.com
git config user.name smoke
mkdir src
for i in 1 2 3; do
    printf 'def helper_%s(value):\n    """Add %s."""\n    return value + %s\n\n\ndef caller_%s():\n    return helper_%s(1)\n' \
        "$i" "$i" "$i" "$i" "$i" > "src/m$i.py"
done
printf 'from src.m1 import helper_1\n\nprint(helper_1(2))\n' > main.py
printf '# user instruction\n' > AGENTS.md
cp AGENTS.md /evidence/original-AGENTS.md
git add -A
git commit -qm init
cp -R "$HOME/project" "$HOME/doctor-project"
pixel setup --json > /evidence/install-1.json
report_ok /evidence/install-1.json
test -s "$HOME/.local/share/pixel/agent-prompt.md"
cp "$HOME/.local/share/pixel/agent-prompt.md" /evidence/first-prompt.md
managed_state > /evidence/managed-1.txt
personal_kept installed
echo 'PASS global install keeps personal settings, hooks and instructions'
pixel setup --json > /evidence/install-2.json
report_ok /evidence/install-2.json
managed_state > /evidence/managed-2.txt
cmp /evidence/managed-1.txt /evidence/managed-2.txt
echo 'PASS global reinstall leaves every managed file unchanged'
# Read the persisted value without the runner's metrics override.
env -u PIXEL_METRICS pixel config metrics off --global
env -u PIXEL_METRICS pixel config metrics > /evidence/metrics.txt
grep -q '^metrics: off' /evidence/metrics.txt
test -s "$HOME/.pixel/config.yaml"
echo 'PASS global config persistence'
test ! -e .pixel/graph.db
start=$(date +%s%N)
pixel audit > /evidence/audit-1.out 2> /evidence/audit-1.err
echo "audit-1-ms: $(( ($(date +%s%N) - start) / 1000000 ))" > /evidence/audit-timing.txt
grep -qx 'pixel: audit: no code graph yet, building it (first run only)' /evidence/audit-1.err
grep -q '^indexed: python 4/4 ' /evidence/audit-1.out
grep -q '^total, 3 of 4 indexed source files' /evidence/audit-1.out
pixel audit > /evidence/audit-2.out 2> /evidence/audit-2.err
if grep -q 'building it' /evidence/audit-2.err; then
    echo 'FAIL second audit rebuilt the graph' >&2
    exit 1
fi
echo 'PASS first audit announces the graph build and reports coverage; second reuses it'
cd "$HOME/doctor-project"
if pixel doctor . --fail-on yellow > /evidence/doctor-before.txt 2>&1; then
    echo 'FAIL doctor passed in a repository that was never prepared' >&2
    exit 1
fi
reds_have_fixes /evidence/doctor-before.txt
pixel doctor . --fix --fail-on red > /evidence/doctor-fix.txt 2>&1
pixel doctor . --fail-on red > /evidence/doctor-after.txt 2>&1
# After --fix only the Codex hook review may stay yellow: no command can
# review a hook for the user, so it must say which step does.
python3 - /evidence/doctor-after.txt <<'PY'
import sys
lines = open(sys.argv[1]).read().splitlines()
yellow = [line.strip() for line in lines if line.lstrip().startswith("[yellow]")]
allowed = ("[yellow] install.codex-hook-review:", "[yellow] repo.codex-hook-review:")
for line in yellow:
    assert line.startswith(allowed), line
    assert "`/hooks`" in line, line
PY
if grep -q 'codex-hook-review' /evidence/doctor-after.txt; then
    echo 'NOTE doctor reports Pixel hooks Codex has not reviewed, naming /hooks'
fi
echo 'PASS doctor names a fix for each red check, and --fix repairs them'
cd "$HOME/project"
pixel setup --repo . --json > /evidence/repo-install.json
report_ok /evidence/repo-install.json
grep -q '^# user instruction$' AGENTS.md
pixel setup --repo . --json > /evidence/repo-install-2.json
report_ok /evidence/repo-install-2.json
echo 'PASS project install preserves user instructions, reinstall succeeds'
# A personal Bash hook holds the Claude guard back: install says what the
# user can do, and doctor stays yellow with no fix line, since no command
# can choose between their hook and the guard.
cp -R "$HOME/doctor-project" "$HOME/guard-project"
cp "$HOME/.claude/settings.json" /evidence/settings-before-bash-hook.json
python3 - "$HOME/.claude/settings.json" <<'PY'
import json
import sys
path = sys.argv[1]
data = json.load(open(path))
data["hooks"]["PreToolUse"].append(
    {"matcher": "Bash", "hooks": [{"type": "command", "command": "/usr/local/bin/my-guard"}]})
json.dump(data, open(path, "w"), indent=2)
PY
(cd "$HOME/guard-project" && pixel setup --repo . --json > /evidence/guard-install.json)
report_ok /evidence/guard-install.json
(cd "$HOME/guard-project" && pixel doctor . --only repo.claude-hooks --fail-on red > /evidence/guard-doctor.txt 2>&1)
python3 - /evidence/guard-install.json /evidence/guard-doctor.txt <<'PY'
import json
import sys
steps = {s["id"]: s for s in json.load(open(sys.argv[1]))["steps"]}
step = steps["hooks.claude"]
assert step["status"] == "yellow", step
summary = step["summary"]
assert "`/usr/local/bin/my-guard`" in summary and "also rewrites shell calls" in summary, summary
if "narrow that hook's `matcher`" not in summary:
    print("NOTE this pixel does not say how to get the guard past a personal Bash hook (AG-05)")
    sys.exit(0)
lines = open(sys.argv[2]).read().splitlines()
yellow = [i for i, line in enumerate(lines) if line.lstrip().startswith("[yellow] repo.claude-hooks:")]
assert len(yellow) == 1, lines
assert "narrow that hook's `matcher`" in lines[yellow[0]], lines
following = lines[yellow[0] + 1] if yellow[0] + 1 < len(lines) else ""
assert not following.lstrip().startswith("fix: "), lines
print("PASS a personal Bash hook holding the guard back is named with what to do, by install and doctor")
PY
cp /evidence/settings-before-bash-hook.json "$HOME/.claude/settings.json"
pixel config classify off
if pixel classify test --label yes --label no > /evidence/classify.out 2>/evidence/classify.err; then
    echo 'FAIL disabled classify succeeded' >&2
    exit 1
fi
test ! -s /evidence/classify.out
grep -q 'classify is disabled' /evidence/classify.err
echo 'PASS disabled classify refuses without a model'
pixel setup --remove --repo . --json > /evidence/repo-uninstall.json
report_ok /evidence/repo-uninstall.json
cmp /evidence/original-AGENTS.md AGENTS.md
echo 'PASS project uninstall restores user instructions exactly'
# Uninstall deletes a binary pixel owns; a copy runs the second check.
cp "$PIXEL_BIN_DIR/pixel" /tmp/pixel-runner
pixel setup --remove --json > /evidence/uninstall-1.json
report_ok /evidence/uninstall-1.json
test ! -e "$HOME/.local/share/pixel/agent-prompt.md"
personal_kept uninstalled
if [ "$PIXEL_OWNS_BINARY" = 1 ]; then
    test ! -e "$PIXEL_BIN_DIR/pixel"
    /tmp/pixel-runner setup --remove --binary-path "$PIXEL_BIN_DIR/pixel" --json > /evidence/uninstall-2.json
else
    # A package manager's binary stays for that manager to remove.
    test -x "$PIXEL_BIN_DIR/pixel"
    pixel setup --remove --json > /evidence/uninstall-2.json
    test -x "$PIXEL_BIN_DIR/pixel"
fi
report_ok /evidence/uninstall-2.json
personal_kept uninstalled
echo 'PASS global uninstall restores personal settings and hooks; repeat succeeds'
# Files left behind are reported, not failed: the contract above is what must hold.
# Shell profile backups sit at the root of the home.
(cd "$HOME" && { find .claude .codex .pi .local/share/pixel .pixel -type f 2>/dev/null;
    find . -maxdepth 1 -type f -name '*.pixel-bak.*' | sed 's|^\./||'; } | sed 's|^|./|' | sort) \
    | comm -13 /evidence/personal-files.txt - > /evidence/residue.txt
echo "NOTE $(wc -l < /evidence/residue.txt) file(s) left after uninstall, listed in residue.txt"
# The backups among them must be exactly those the last uninstall reported,
# with a removal command a shell reads back as the same paths; the rest is
# the pixel config the checks set with `pixel config --global`.
python3 - /evidence/uninstall-2.json /evidence/residue.txt <<'PY'
import json
import os
import shlex
import sys
report, residue = sys.argv[1], sys.argv[2]
home = os.path.expanduser("~")
steps = [s for s in json.load(open(report))["steps"] if s["id"] == "backups"]
left = open(residue).read().splitlines()
if not steps:
    print("NOTE this pixel does not report the backups uninstall leaves (IN-03)")
    sys.exit(0)
detail = steps[0].get("detail")
words = shlex.split(detail) if detail else ["rm", "--"]
assert words[:2] == ["rm", "--"], detail
reported = sorted("./" + os.path.relpath(p, home) for p in words[2:])
backups = sorted(line for line in left if ".pixel-bak." in line)
assert reported == backups, (reported, backups)
others = sorted(line for line in left if ".pixel-bak." not in line)
assert others == ["./.pixel/config.yaml"], others
print(f"PASS uninstall reports the {len(reported)} backup(s) left, and how to remove them")
PY
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 466kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 466kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 466kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 192kUpdated Claude CodeCodexVerifiedInstall