ai-cost-guard

When to invoke

Use when: "ai cost guard", "cap llm spend", "unbounded llm loop", "token budget", "runaway api cost", "bedrock spend limit", "why was the Anthropic/OpenAI bill so high", or before merging a change that adds a call to a paid model API.

Preamble

eval "$(~/.vibestack/bin/vibe-slug 2>/dev/null)" 2>/dev/null || SLUG="unknown"
_LEARN_FILE="${VIBESTACK_HOME:-$HOME/.vibestack}/projects/${SLUG:-unknown}/learnings.jsonl"
if [ -f "$_LEARN_FILE" ]; then
  _LEARN_COUNT=$(wc -l < "$_LEARN_FILE" 2>/dev/null | tr -d ' ')
  echo "LEARNINGS: $_LEARN_COUNT entries loaded"
  if [ "$_LEARN_COUNT" -gt 5 ] 2>/dev/null; then
    ~/.vibestack/bin/vibe-learnings-search --limit 5 2>/dev/null || true
  fi
else
  echo "LEARNINGS: none yet"
fi

{{include lib/snippets/session-host.md}}

{{include lib/snippets/decision-brief.md}}

{{include lib/snippets/working-protocols.md}}

{{include lib/snippets/state-protocols.md}}

User-invocable

When the user types /ai-cost-guard, run this skill.

Step 1: Inventory paid-API call sites

Every call to a metered model API is a place where money leaves the account. Find all of them before judging any of them.

grep -rniE \
  'anthropic|@anthropic-ai/sdk|openai|bedrock-runtime|invoke_?model|converse|boto3.*bedrock|google\.generativeai|replicate|elevenlabs|deepgram|fal\.ai|fal_client|@fal-ai|messages\.create|chat\.completions\.create|responses\.create|generate_content|replicate\.run|\.transcribe|\.subscribe' \
  --include='*.py' --include='*.ts' --include='*.tsx' --include='*.js' --include='*.mjs' \
  --include='*.go' --include='*.rb' --include='*.java' --include='*.kt' --include='*.rs' \
  --exclude-dir=node_modules --exclude-dir=.git --exclude-dir=dist --exclude-dir=build \
  --exclude-dir=vendor --exclude-dir=.venv . 2>/dev/null

The -i matters: a class name (Anthropic(), new OpenAI()) and a boto3 method (invoke_model, converse) differ in case from the package and API-operation spellings.