Audits and rewrites your OpenClaw config to minimize API token costs. Use this skill whenever the user mentions high bills, API costs, billing changes, "too...
Security Analysis
high confidenceThe skill's requests and instructions are consistent with its stated purpose (auditing and rewriting OpenClaw configs to reduce Anthropic API spend); it reads and may write local OpenClaw config files and uses an internal pricing reference, and it does not request unrelated credentials or download external code.
Name/description match the behavior: the skill inspects OpenClaw YAML/JSON, applies cost-focused edits (caching, model routing, thinking scope, context1m, fastMode) and estimates savings using a local pricing reference. It does not request unrelated secrets or binaries.
Instructions explicitly direct the agent to locate and read OpenClaw config files from the current directory and several home paths, analyze them using the bundled pricing reference, present a diff and cost estimate, then (with user confirmation) write the optimized config back to the same file. This is within the stated purpose but involves reading and modifying user files — the skill does ask for confirmation before writing, which mitigates risk, but users should review diffs and back up configs before applying changes.
No install spec and no code files — instruction-only skill. Lowest risk from installation because nothing is downloaded or installed on disk beyond the agent's normal execution environment.
No environment variables, credentials, or external config paths are requested. The skill mentions that caching is API-key only, but it does not request API keys itself — this is consistent (it can only recommend config changes; applying caching may require the user to use API keys).
Skill is user-invocable and not forced-always; model invocation is allowed (normal default). Because it can write to local config files (with confirmation), consider that an agent with file-system access could apply changes. There's no evidence the skill modifies other skills or system-wide agent settings.
Guidance
This skill appears to do what it claims, but you should: (1) back up your OpenClaw config before running it, (2) review the reported issues, estimated savings, and the exact diff it proposes before approving any write, (3) be aware caching suggestions may require API-key usage (subscriptions/setup-tokens may not honor caching), and (4) only allow the skill to run with a user who understands and can validate the changes. Because the skill will read and (with confirmation) write files in your home/project paths, treat it like any tool that edits configuration — inspect outputs and keep a copy of the original config.
Latest Release
v1.0.0
anthropic-cost-optimizer 1.0.0 — initial release - Audits OpenClaw configuration for Anthropic API cost-saving opportunities. - Identifies and fixes issues in five main cost levers: prompt caching, model routing, thinking level, 1M context window, and fast mode. - Flags each issue by impact severity (HIGH, MEDIUM, LOW) and recommends targeted optimizations. - Generates annotated config diffs and estimates monthly savings before applying changes. - Handles YAML/JSON configs, edge cases, and provides clear user prompts for confirmation and next steps.
Popular Skills
Published by @nextfrontierbuilds on ClawHub
