Anthropic Token Optimizer

v1.3.0

Reduce Anthropic API costs (cache read, compaction, context bloat) for OpenClaw agents. Use when users ask about token optimization, reducing API costs, cach...

0· 42·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
Name and description claim token-cost/compaction/workspace guidance and the SKILL.md contains only guidance and concrete config/file-management suggestions that match that purpose. No unrelated credentials, binaries, or system-level access are requested.
Instruction Scope
Instructions focus on config changes (openclaw.json), workspace hygiene, compaction/heartbeat patterns, and creating/maintaining files (MEMORY.md, SESSION-STATE.md, atris/MAP.md). This is within scope, but the doc also suggests running tools/commands (npx, rg, clawhub, grep) and using filesystem commands like `/context list` which are not declared in the registry metadata — users should expect the agent or operator to run these or ensure those binaries exist. The skill will recommend writing and reading local workspace files (intended behavior) — verify no sensitive secrets will be persisted accidentally.
Install Mechanism
There is no install spec in the registry (instruction-only). The SKILL.md suggests a manual npx-based installation step for the optional 'atris' pattern (npx clawhub@latest install atris). That is a manual command that would fetch code from npm when run; the skill itself does not automatically download or install anything. Users should be aware that following that suggestion will run a network-installed package.
Credentials
The skill requests no environment variables, credentials, or config paths. All recommended actions are local config edits and file hygiene; the lack of requested secrets is proportional to the described functionality.
Persistence & Privilege
Skill is user-invocable, not always-on, and does not request persistent privileges or modify other skills. It advises creating/maintaining local files but does not itself persist or alter agent-wide settings via an install spec.
Assessment
This is an instruction-only optimization guide (no code shipped, no credentials requested). It's coherent with its purpose, but before applying recommendations: (1) back up openclaw.json and workspace files before making config changes; (2) verify Anthropic pricing/metrics against official docs—numbers in the guide may be estimates; (3) confirm availability of any referenced tools (npx, rg/ripgrep, clawhub) before running suggested commands; (4) be cautious about persisting sensitive data to MEMORY.md or MAP.md (avoid storing PII/credentials); and (5) note that following the npx/clawhub suggestion will install third-party code from npm — inspect that package before running it.

Like a lobster shell, security has layers — review code before you run it.

anthropicvk97ac6898x7pxrsg50stkz74rd843zsfcachevk97ac6898x7pxrsg50stkz74rd843zsfcodebase-mapvk97ac6898x7pxrsg50stkz74rd843zsfcontext-managementvk97ac6898x7pxrsg50stkz74rd843zsfcost-reductionvk97ac6898x7pxrsg50stkz74rd843zsflatestvk97ac6898x7pxrsg50stkz74rd843zsftoken-optimizationvk97ac6898x7pxrsg50stkz74rd843zsf

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Comments