flurdy/agent-skills / skills/model-update-check
model-update-check
VerifiedAudit Pi startup defaults, model routing, discovery and coordinated migration previews across second-opinion and pi-spend; read-only by default, with separately confirmed native catalog refresh.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 21.6 KB |
| references/companion-apply.md | markdown | 5.8 KB |
| references/migration-preview.md | markdown | 10.9 KB |
| scripts/apply_migration.py | script | 25.6 KB |
| scripts/catalog_refresh.py | script | 5.8 KB |
| scripts/claude_discovery.py | script | 4.9 KB |
| scripts/migration_audit.py | script | 43.9 KB |
| scripts/model-update-check.shexecutable | script | 20.6 KB |
| scripts/startup_audit.py | script | 8.2 KB |
| tests/test-model-update-check.shexecutable | script | 21.8 KB |
| tests/test_apply_migration.py | script | 19.1 KB |
| tests/test_claude_discovery.py | script | 5.9 KB |
| tests/test_migration_audit.py | script | 39.6 KB |
| tests/test_startup_audit.py | script | 12.6 KB |
tests/test-model-update-check.sh
View on GitHub#!/usr/bin/env bash
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
HELPER="$(cd "$SCRIPT_DIR/.." && pwd)/scripts/model-update-check.sh"
ORIGINAL_PATH="$PATH"
fail() {
printf 'FAIL: %s\n' "$*" >&2
exit 1
}
TMP_DIR="$(mktemp -d)"
trap 'rm -rf "$TMP_DIR"' EXIT
mkdir -p "$TMP_DIR/bin" "$TMP_DIR/home/.pi/agent"
printf '{}\n' > "$TMP_DIR/home/.pi/agent/auth.json"
PI_LOG="$TMP_DIR/pi.log"
CURL_LOG="$TMP_DIR/curl.log"
BREW_LOG="$TMP_DIR/brew.log"
export PI_LOG CURL_LOG BREW_LOG
cat > "$TMP_DIR/bin/pi" <<'FAKE_PI'
#!/usr/bin/env bash
set -euo pipefail
printf '%s\n' "$*" >> "$PI_LOG"
case "$*" in
--version)
printf '%s\n' '0.80.0'
;;
update\ --help)
[[ "${REFRESH_CASE:-ok}" != unsupported ]] || exit 0
printf ' --models Refresh model catalogs only\n'
;;
update\ --models)
[[ "${REFRESH_CASE:-ok}" != fail ]] || exit 1
if [[ "${REFRESH_CASE:-ok}" == timeout ]]; then sleep 5; fi
printf 'native-output-must-not-be-exposed\n'
;;
--offline\ --no-extensions\ --no-skills\ --no-prompt-templates\ --no-themes\ --no-context-files\ --no-approve\ --list-models)
cat <<'MODELS'
provider model context max-out thinking images
anthropic claude-sonnet-5 1M 128K yes yes
openai-codex gpt-5.6-terra 372K 128K yes yes
google gemini-3.5-flash 1M 64K yes yes
openrouter qwen/current-reasoner 262K 32K yes no
openrouter ~google/gemini-flash-latest 1M 64K yes yes
MODELS
;;
*)
printf 'unexpected fake pi arguments: %s\n' "$*" >&2
exit 2
;;
esac
FAKE_PI
chmod +x "$TMP_DIR/bin/pi"
cat > "$TMP_DIR/bin/claude" <<'FAKE_CLAUDE'
#!/usr/bin/env bash
set -euo pipefail
[[ "$*" == --version ]] || exit 2
printf '2.1.499 (Claude Code)\n'
FAKE_CLAUDE
chmod +x "$TMP_DIR/bin/claude"
cat > "$TMP_DIR/bin/brew" <<'FAKE_BREW'
#!/usr/bin/env bash
set -euo pipefail
printf '%s\n' "$*" >> "$BREW_LOG"
[[ "$1" == "info" && "$2" == "--json=v2" && "$3" == "pi-coding-agent" ]] || exit 2
outdated="${BREW_OUTDATED:-true}"
cat <<JSON
{"formulae":[{"name":"pi-coding-agent","installed":[{"version":"0.80.0"}],"versions":{"stable":"0.81.0"},"outdated":$outdated}]}
JSON
FAKE_BREW
chmod +x "$TMP_DIR/bin/brew"
cat > "$TMP_DIR/models-dev.json" <<'JSON'
{
"anthropic": {
"models": {
"claude-sonnet-5": {
"name": "Claude Sonnet 5",
"release_date": "2026-06-29",
"reasoning": true,
"limit": {"context": 1000000, "output": 128000}
}
}
},
"openai": {
"models": {
"gpt-5.6-terra": {
"name": "GPT-5.6 Terra",
"release_date": "2026-07-09",
"reasoning": true,
"limit": {"context": 372000, "output": 128000}
}
}
},
"google": {
"models": {
"gemini-3.5-flash": {
"name": "Gemini 3.5 Flash",
"release_date": "2026-05-19",
"reasoning": true
}
}
},
"openrouter": {
"models": {
"qwen/new-reasoner": {
"name": "New Qwen Reasoner",
"release_date": "2026-07-01",
"reasoning": true
},
"x-ai/current-critic": {
"name": "Current xAI Critic",
"release_date": "2026-07-08",
"reasoning": true
},
"~google/gemini-flash-latest": {
"name": "Gemini Flash Latest",
"release_date": "2026-04-27",
"reasoning": true
}
}
}
}
JSON
cat > "$TMP_DIR/bin/curl" <<'FAKE_CURL'
#!/usr/bin/env bash
set -euo pipefail
url=""
output=""
original_args="$*"
while [[ $# -gt 0 ]]; do
case "$1" in
--output)
output="$2"
shift 2
;;
http://*|https://*)
url="$1"
shift
;;
*)
shift
;;
esac
done
[[ -n "$url" && -n "$output" ]] || exit 2
printf '%s\n' "$url" >> "$CURL_LOG"
case "$url" in
https://fixture.test/models)
cp "$FIXTURE_MODELS_DEV" "$output"
;;
https://fixture.test/openrouter)
cat > "$output" <<'JSON'
{"data":[
{"id":"qwen/current-reasoner","name":"Current Qwen","created":1,"context_length":262144,"pricing":{},"expiration_date":null},
{"id":"x-ai/current-critic","name":"Current xAI","created":2,"context_length":500000,"pricing":{},"expiration_date":null},
{"id":"~google/gemini-flash-latest","name":"Gemini Flash Latest","created":3,"context_length":1048576,"pricing":{},"expiration_date":null}
]}
JSON
;;
https://code.claude.com/docs/en/model-config.md)
[[ "$original_args" != *--location* && "$original_args" == *"--proto =https"* ]] || exit 2
cat > "$output" <<'DOCS'
### Model aliases
| Model alias | Behavior |
| **`sonnet`** | Latest Sonnet |
| **`opus`** | Latest Opus |
| **`fable`** | Provider-dependent Fable |
### Versions
Sonnet 9.1 requires Claude Code v2.1.500 or later.
DOCS
;;
https://fixture.test/pi-latest)
printf '%s\n' '{"version":"0.81.0"}' > "$output"
;;
*)
exit 22
;;
esac
FAKE_CURL
chmod +x "$TMP_DIR/bin/curl"
ROUTER_CONFIG="$TMP_DIR/router.json"
cat > "$ROUTER_CONFIG" <<'JSON'
{
"enabled": true,
"tiers": {
"standard": {
"candidates": [
{"model": "openai-codex/gpt-5.6-terra", "metered": false},
{"model": "anthropic/claude-sonnet-5", "metered": true}
]
},
"economy": {
"candidates": [
{"model": "google/gemini-3.5-flash", "metered": true},
{"model": "openrouter/~google/gemini-flash-latest", "metered": true},
{"model": "custom/example-model", "metered": true}
]
}
}
}
JSON
CONSENSUS_CONFIG="$TMP_DIR/consensus.json"
cat > "$CONSENSUS_CONFIG" <<'JSON'
{
"version": 1,
"profiles": {
"test": {
"models": [
{"model": "openrouter/qwen/current-reasoner", "vendor": "Qwen", "role": "reasoning"},
{"model": "openrouter/x-ai/current-critic", "vendor": "xAI", "role": "critique"}
],
"limits": {
"maxParallel": 2,
"maxPromptBytes": 1024,
"maxOutputTokensPerModel": 100,
"defaultTimeoutSeconds": 5
}
},
"hybrid": {
"quorum": 3,
"consensusQuorum": 4,
"routes": [
{"id": "claude", "kind": "local", "agent": "claude", "model": "fable", "effort": "high", "role": "reasoning"},
{"id": "codex", "kind": "local", "agent": "codex", "role": "critique"},
{"id": "qwen", "kind": "openrouter", "model": "openrouter/qwen/current-reasoner", "vendor": "Qwen", "role": "verification", "effort": "high"},
{"id": "xai", "kind": "openrouter", "model": "openrouter/x-ai/current-critic", "vendor": "xAI", "role": "adversarial review", "maxOutputTokens": 2000},
{"id": "gemini", "kind": "openrouter", "model": "openrouter/~google/gemini-flash-latest", "vendor": "Google", "role": "fast critique"}
],
"limits": {
"maxParallel": 4,
"maxPromptBytes": 65536,
"maxOutputTokensPerModel": 16000,
"defaultTimeoutSeconds": 1800
}
}
}
}
JSON
export FIXTURE_MODELS_DEV="$TMP_DIR/models-dev.json"
RUN_ENV=(env \
"PATH=$TMP_DIR/bin:$ORIGINAL_PATH" \
"HOME=$TMP_DIR/home" \
"PI_CODING_AGENT_DIR=$TMP_DIR/home/.pi/agent" \
"MODELS_DEV_URL=https://fixture.test/models" \
"OPENROUTER_MODELS_URL=https://fixture.test/openrouter" \
"PI_PACKAGE_URL=https://fixture.test/pi-latest")
COMMON_ARGS=(--router-config "$ROUTER_CONFIG" --consensus-config "$CONSENSUS_CONFIG")
result_json="$("${RUN_ENV[@]}" "$HELPER" "${COMMON_ARGS[@]}")"
jq -e '
.mode == "hybrid" and
.readOnly == true and
.schemaVersion == 2 and
.refresh.status == "not-requested" and
any(.configurationInventory[];
.identity == "local/claude/fable" and .resolution == "native-alias") and
any(.configurationInventory[];
.identity == "openai-codex/gpt-5.6-terra" and .metered == false) and
.sources.routerConfig.status == "ok" and
.sources.consensusConfig.status == "ok" and
.sources.piCatalog.installedVersion == "0.80.0" and
.sources.piRelease.latestVersion == "0.81.0" and
.sources.homebrew.status == "ok" and
.sources.homebrew.formula == "pi-coding-agent" and
.sources.homebrew.installedVersion == "0.80.0" and
.sources.homebrew.latestVersion == "0.81.0" and
.piUpdateAvailable == true and
.piNpmUpdateAvailable == true and
.piHomebrewUpdateAvailable == true and
(.configuredModels | length == 10) and
any(.configuredModels[];
.model == "openai-codex/gpt-5.6-terra" and
.catalogProvider == "openai" and .piAvailable == true and .liveFound == true) and
any(.configuredModels[];
.model == "custom/example-model" and
.catalogProviderFound == false and .liveFound == null) and
any(.configuredModels[];
.model == "openrouter/x-ai/current-critic" and
.piAvailable == false and .liveFound == true and .openRouterFound == true) and
any(.configuredModels[];
.usage == "test" and .model == "openrouter/qwen/current-reasoner" and
.piAvailable == true and .liveFound == false) and
any(.configuredModels[];
.usage == "hybrid" and .model == "openrouter/x-ai/current-critic" and
.piAvailable == false and .liveFound == true and .openRouterFound == true) and
any(.configuredModels[];
.model == "openrouter/~google/gemini-flash-latest" and
.catalogModel == "~google/gemini-flash-latest" and
.piAvailable == true and .liveFound == true and .openRouterFound == true) and
all(.configuredModels[]; .model != "fable" and .model != "native-default") and
any(.findings[]; .kind == "pi-update" and .manager == "homebrew") and
any(.findings[]; .kind == "pi-unavailable" and .model == "openrouter/x-ai/current-critic") and
any(.findings[]; .kind == "live-missing" and .model == "openrouter/qwen/current-reasoner") and
all(.findings[]; .kind != "openrouter-missing" or .model != "openrouter/~google/gemini-flash-latest") and
(.recentOpenRouterByNamespace.qwen[0].model == "qwen/new-reasoner") and
(.recentOpenRouterByNamespace | has("google") and (has("~google") | not))
' <<< "$result_json" >/dev/null || fail "hybrid audit output was incorrect"
[[ "$(wc -l < "$CURL_LOG" | tr -d '[:space:]')" -eq 4 ]] || \
fail "hybrid mode did not make exactly four public metadata requests"
jq -e '.sources.claudeClient.installedVersion == "2.1.499" and
.sources.claudeDocumentation.status == "ok" and
.claudeDiscovery.inventory.status == "not-assessed" and
.claudeDiscovery.accountAccess == "unverified"' <<< "$result_json" >/dev/null || fail "Claude fallback provenance"
grep -Fqx -- 'info --json=v2 pi-coding-agent' "$BREW_LOG" || \
fail "hybrid mode did not inspect Homebrew formula metadata"
result_json="$(BREW_OUTDATED=false "${RUN_ENV[@]}" "$HELPER" "${COMMON_ARGS[@]}")"
jq -e '
.sources.homebrew.status == "ok" and
.piNpmUpdateAvailable == true and
.piHomebrewUpdateAvailable == false and
.piUpdateAvailable == false and
any(.findings[]; .kind == "pi-npm-ahead-of-homebrew") and
all(.findings[]; .kind != "pi-update")
' <<< "$result_json" >/dev/null || fail "Homebrew state did not override npm update availability"
CLAUDE_ROUTER="$TMP_DIR/claude-router.json"
printf '%s\n' '{"tiers":{"standard":{"candidates":[
{"model":"pi-claude-cli/claude-sonnet-9-1","metered":false},
{"model":"pi-claude-cli/opus","metered":false},
{"model":"pi-claude-cli/claude-haiku-9-1","metered":false,"enabled":false}
]}}}' > "$CLAUDE_ROUTER"
result_json="$(BREW_OUTDATED=false "${RUN_ENV[@]}" "$HELPER" --router-config "$CLAUDE_ROUTER" --consensus-config "$CONSENSUS_CONFIG")"
jq -e '
all(.configuredModels[] | select(.model | startswith("pi-claude-cli/"));
.piAvailable == null and .piAssessment == "out-of-scope") and
all(.findings[]; ((.kind != "pi-unavailable" and .kind != "live-missing") or (.model | startswith("pi-claude-cli/") | not))) and
any(.configurationInventory[]; .identity == "pi-claude-cli/claude-sonnet-9-1" and
.claudeClientSupport.status == "minimum-version-mismatch") and
any(.configurationInventory[]; .identity == "pi-claude-cli/opus" and .resolution == "native-alias") and
any(.configurationInventory[]; .identity == "pi-claude-cli/claude-haiku-9-1" and .enabled == false) and
.verdict == "INCOMPLETE EVIDENCE"
' <<< "$result_json" >/dev/null || fail "extension-disabled catalogue produced false model availability"
: > "$CURL_LOG"
: > "$PI_LOG"
result_json="$("${RUN_ENV[@]}" "$HELPER" --offline "${COMMON_ARGS[@]}")"
jq -e '
.mode == "offline" and
.sources.modelsDev.status == "skipped" and
.sources.openRouter.status == "skipped" and
.sources.piRelease.status == "skipped" and
.sources.claudeDocumentation.status == "skipped" and
.sources.homebrew.status == "ok" and
.piUpdateAvailable == true and
.piNpmUpdateAvailable == false and
.piHomebrewUpdateAvailable == true and
all(.configuredModels[]; .liveFound == null)
' <<< "$result_json" >/dev/null || fail "offline audit output was incorrect"
[[ ! -s "$CURL_LOG" ]] || fail "offline mode invoked curl"
grep -Fqx -- '--offline --no-extensions --no-skills --no-prompt-templates --no-themes --no-context-files --no-approve --list-models' "$PI_LOG" || \
fail "offline mode did not constrain Pi startup"
# Startup sources are optional by default, explicit missing paths are actionable.
jq -e '.sources.launcherConfig.status == "missing" and .sources.piSettings.status == "missing" and
.startupAudit.auditOnly and .startupAudit.findings == []' <<< "$result_json" >/dev/null || fail "missing optional startup sources"
LAUNCHER="$TMP_DIR/home/.pi/agent/pl-launcher.json"
SETTINGS="$TMP_DIR/home/.pi/agent/settings.json"
printf '%s\n' '{"defaultModel":"openai-codex/example-old","defaultThinking":"high"}' > "$LAUNCHER"
printf '%s\n' '{"defaultProvider":"openai-codex","defaultModel":"example-other","enabledModels":["openai-codex/gpt-5.6-terra"],"apiKey":"PRIVATE_SENTINEL"}' > "$SETTINGS"
startup_before="$(sha256sum "$LAUNCHER" "$SETTINGS")"
result_json="$("${RUN_ENV[@]}" "$HELPER" --offline "${COMMON_ARGS[@]}")"
jq -e --arg launcher "$LAUNCHER" --arg settings "$SETTINGS" '
.sources.launcherConfig.path == $launcher and .sources.piSettings.path == $settings and
.sources.launcherConfig.status == "ok" and .sources.piSettings.status == "ok" and
(.sources.piSettings.sha256 | length == 64) and
.startupAudit.defaults.piSettings.enabledModelsMembership == "not-exactly-listed" and
any(.startupAudit.findings[]; .code == "startup-defaults-disagree") and
.interaction.primaryAction.kind == "inspect-startup" and
([.startupAudit.launchPaths[].path] == ["fresh-pl","plain-pi","resume-continue","orca"]) and
.recommendations == []' <<< "$result_json" >/dev/null || fail "startup defaults or provenance"
[[ "$result_json" != *PRIVATE_SENTINEL* ]] || fail "unrelated settings leaked"
[[ "$startup_before" == "$(sha256sum "$LAUNCHER" "$SETTINGS")" ]] || fail "startup sources mutated"
# The Pi agent-dir override affects settings, not the launcher's fixed home path.
mkdir -p "$TMP_DIR/alternate-agent"
printf '{}\n' > "$TMP_DIR/alternate-agent/settings.json"
result_json="$("${RUN_ENV[@]}" PI_CODING_AGENT_DIR="$TMP_DIR/alternate-agent" "$HELPER" --offline "${COMMON_ARGS[@]}")"
jq -e --arg path "$TMP_DIR/alternate-agent/settings.json" --arg launcher "$LAUNCHER" '
.sources.piSettings.path == $path and .sources.launcherConfig.path == $launcher
' <<< "$result_json" >/dev/null || fail "agent-dir settings path"
result_json="$("${RUN_ENV[@]}" "$HELPER" --offline "${COMMON_ARGS[@]}" --launcher-config "$TMP_DIR/missing-launcher" --settings-config "$TMP_DIR/missing-settings")"
jq -e '.sources.launcherConfig.required and .sources.piSettings.required and
.sources.piSettings.status == "missing" and
([.startupAudit.findings[] | select(.code == "startup-config-unavailable")] | length == 2)
' <<< "$result_json" >/dev/null || fail "explicit missing startup sources"
rm -f "$LAUNCHER" "$SETTINGS"
INVALID_ROUTER="$TMP_DIR/invalid-router.json"
printf '%s\n' '{}' > "$INVALID_ROUTER"
result_json="$("${RUN_ENV[@]}" "$HELPER" --offline \
--router-config "$INVALID_ROUTER" --consensus-config "$CONSENSUS_CONFIG")"
jq -e '
.sources.routerConfig.status == "invalid" and
(.configuredModels | length == 5) and
any(.findings[]; .kind == "config" and .source == "model-tier-router")
' <<< "$result_json" >/dev/null || fail "invalid config did not degrade independently"
assert_invalid_panel() {
local fixture="$1"
local label="$2"
result_json="$("${RUN_ENV[@]}" "$HELPER" --offline \
--router-config "$ROUTER_CONFIG" --consensus-config "$fixture")"
jq -e . <<< "$result_json" >/dev/null || fail "$label did not emit JSON"
jq -e '
.sources.routerConfig.status == "ok" and
.sources.consensusConfig.status == "invalid" and
(.configuredModels | length == 5) and
any(.findings[]; .kind == "config" and .source == "second-opinion" and
.message == "panel config is invalid")
' <<< "$result_json" >/dev/null || fail "$label did not degrade independently"
}
for budget in 64000 128000; do
LARGE_OUTPUT_PANEL="$TMP_DIR/large-output-$budget.json"
jq --argjson budget "$budget" '
.profiles.test.limits.maxOutputTokensPerModel = $budget |
.profiles.test.models[0].maxOutputTokens = $budget |
.profiles.hybrid.limits.maxOutputTokensPerModel = $budget |
.profiles.hybrid.routes[3].maxOutputTokens = $budget
' "$CONSENSUS_CONFIG" > "$LARGE_OUTPUT_PANEL"
result_json="$("${RUN_ENV[@]}" "$HELPER" --offline \
--router-config "$ROUTER_CONFIG" --consensus-config "$LARGE_OUTPUT_PANEL")"
jq -e '
.sources.consensusConfig.status == "ok" and
(.configuredModels | length == 10) and
all(.findings[]; .kind != "config" or .source != "second-opinion")
' <<< "$result_json" >/dev/null || fail "supported $budget output budget was rejected"
done
for budget in '128001' '0' '-1' '1.5' '"64000"' 'null' 'true'; do
INVALID_OUTPUT_PANEL="$TMP_DIR/invalid-output.json"
jq --argjson budget "$budget" '.profiles.hybrid.limits.maxOutputTokensPerModel = $budget' \
"$CONSENSUS_CONFIG" > "$INVALID_OUTPUT_PANEL"
assert_invalid_panel "$INVALID_OUTPUT_PANEL" "invalid profile output budget $budget"
done
INVALID_ROUTE_OUTPUT="$TMP_DIR/invalid-route-output.json"
jq '.profiles.hybrid.limits.maxOutputTokensPerModel = 64000 |
.profiles.hybrid.routes[3].maxOutputTokens = 64001' \
"$CONSENSUS_CONFIG" > "$INVALID_ROUTE_OUTPUT"
assert_invalid_panel "$INVALID_ROUTE_OUTPUT" "route output budget above profile ceiling"
jq '.profiles.test.limits.maxOutputTokensPerModel = 64000 |
.profiles.test.models[0].maxOutputTokens = 64001' \
"$CONSENSUS_CONFIG" > "$INVALID_ROUTE_OUTPUT"
assert_invalid_panel "$INVALID_ROUTE_OUTPUT" "legacy model output budget above profile ceiling"
INVALID_PANEL="$TMP_DIR/invalid-panel.json"
jq '.profiles.hybrid.models = .profiles.test.models' "$CONSENSUS_CONFIG" > "$INVALID_PANEL"
assert_invalid_panel "$INVALID_PANEL" "models/routes conflict"
INVALID_GEMINI_EFFORT="$TMP_DIR/invalid-gemini-effort.json"
jq '.profiles.hybrid.routes[1] = {id:"gemini",kind:"local",agent:"gemini",effort:"high",role:"critique"}' \
"$CONSENSUS_CONFIG" > "$INVALID_GEMINI_EFFORT"
assert_invalid_panel "$INVALID_GEMINI_EFFORT" "unsupported Gemini effort"
INVALID_OPENROUTER_ID="$TMP_DIR/invalid-openrouter-id.json"
jq '.profiles.hybrid.routes[2].model = "openrouter//missing-provider"' \
"$CONSENSUS_CONFIG" > "$INVALID_OPENROUTER_ID"
assert_invalid_panel "$INVALID_OPENROUTER_ID" "malformed OpenRouter identity"
INVALID_QUORUM="$TMP_DIR/invalid-quorum.json"
jq '.profiles.hybrid.quorum = 6' "$CONSENSUS_CONFIG" > "$INVALID_QUORUM"
assert_invalid_panel "$INVALID_QUORUM" "quorum above enabled route count"
# Conversational evidence uses the same validation/digests without a user-authored file.
result_json="$(printf '%s' '{"schemaVersion":1,"recommendations":[]}' | "${RUN_ENV[@]}" "$HELPER" --offline "${COMMON_ARGS[@]}" --evidence -)"
jq -e '.sources.releaseEvidence.status == "ok" and .sources.releaseEvidence.path == "stdin" and
(.sources.releaseEvidence.sha256 | length == 64) and .readOnly' <<< "$result_json" >/dev/null || fail "stdin evidence contract"
result_json="$(printf '%s' '{}' | "${RUN_ENV[@]}" "$HELPER" --offline "${COMMON_ARGS[@]}" --evidence -)"
jq -e '.sources.releaseEvidence.status == "invalid-or-stale" and .recommendations == []' <<< "$result_json" >/dev/null || fail "stdin bypassed validation"
if python3 -c 'print("x" * 65537)' | "${RUN_ENV[@]}" "$HELPER" --offline "${COMMON_ARGS[@]}" --evidence - >/dev/null 2>&1; then
fail "oversized stdin evidence accepted"
fi
# Refresh never runs implicitly, and failed authorization never reaches native update.
if grep -q '^update ' "$PI_LOG"; then fail "default/offline invoked update"; fi
for extra in denied conflict unsupported fail timeout ok; do
: > "$PI_LOG"
refresh_flags=(--refresh-models --confirm-refresh)
expected="$extra"
case "$extra" in
denied) refresh_flags=(--refresh-models); expected=authorization-required ;;
conflict) refresh_flags+=(--offline); expected=offline-conflict ;;
timeout) refresh_flags+=(--refresh-timeout 1) ;;
fail) expected=failed ;;
esac
result_json="$(REFRESH_CASE="$extra" "${RUN_ENV[@]}" "$HELPER" "${COMMON_ARGS[@]}" "${refresh_flags[@]}")"
jq -e --arg expected "$expected" '.refresh.status == $expected' <<< "$result_json" >/dev/null || fail "wrong refresh status: $extra"
if [[ "$extra" == ok ]]; then
jq -e '.refresh.fresh and (.readOnly == false)' <<< "$result_json" >/dev/null || fail "refresh success contract"
# Reacquisition must follow the models-only update, never precede it.
first="$(head -n 1 "$PI_LOG")"
[[ "$first" == 'update --help' ]] || fail "refresh capability probe was not first"
[[ "$(grep -n '^update --models$' "$PI_LOG" | cut -d: -f1)" -lt "$(grep -n '^--offline ' "$PI_LOG" | cut -d: -f1)" ]] || fail "catalog was not reacquired"
else
jq -e '.refresh.fresh == false and .verdict != "CURRENT"' <<< "$result_json" >/dev/null || fail "false freshness: $extra"
fi
if [[ "$extra" == denied || "$extra" == conflict ]]; then
if grep -q '^update ' "$PI_LOG"; then fail "unauthorized native command"; fi
fi
if grep -Eq '^update($| --(self|all|extensions)$)' "$PI_LOG"; then fail "package update fallback"; fi
[[ "$result_json" != *native-output-must-not-be-exposed* ]] || fail "native output leaked"
done
printf '%s\n' 'model-update-check tests passed'
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 471k | Updated | Claude CodeCodex | Verified | Install |
| 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: 471k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |