Claude API Cost Optimizer
ReviewAudited by ClawScan on May 6, 2026.
Overview
This instruction-only skill is coherent with its cost-optimization purpose, but users should notice that it uses an Anthropic API key and may create local cache/log files containing prompts, code, or API responses.
This appears safe to review and use as guidance, but treat its cache and log examples carefully: use a dedicated Anthropic API key if possible, avoid caching secrets or personal data, restrict local file permissions, and regularly delete cache/log files.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If used carelessly, API calls may incur costs or use the wrong Anthropic account/key.
The skill requires an Anthropic API key so its examples can call the Claude API; this is purpose-aligned but gives the workflow access to billable account credentials.
Ensure `ANTHROPIC_API_KEY` is set in your environment (required).
Use a scoped key where possible, monitor usage, and avoid exposing the key in logs, prompts, cache files, or shared environments.
Private code, prompts, or API responses could remain on disk and be exposed through backups, repository commits, weak file permissions, or shared machines.
The skill intentionally recommends local caching and cost logging, which can persist sensitive prompt/code/response content across sessions.
This skill creates `~/.claude_cache/`, `costs.log`, and `batch_results.jsonl` which may contain source code, API responses, and prompts.
Follow the checklist: do not cache secrets or PII, restrict permissions, set short TTLs, clear caches regularly, and add cache/log paths to `.gitignore`.