TokenKiller

Reduces token usage across multi-skill agent workflows (search, coding, debugging, testing, docs) using budgets, gating, progressive disclosure, and deduped...

This is an instruction-only policy that appears coherent and low-risk from a supply chain/credential perspective: it asks for nothing sensitive and doesn't install code. Before enabling it, review the SKILL.md to confirm you accept its operational tradeoffs—especially that it will (a) limit reads/outputs, (b) ask at most three clarifying questions then proceed on defaults, and (c) suppress large dumps unless L3 pull scenarios or explicit user request occur. Test it on non-production workflows first to ensure the throttling rules don't cause the agent to take undesired actions or miss important clarifications. If you rely on full logs or exhaustive outputs by default, be aware TokenKiller will suppress those unless explicitly overridden.

Type: OpenClaw Skill Name: tokenkiller Version: 1.0.1 The 'tokenkiller' skill is a utility designed to optimize token consumption and reduce costs in AI agent workflows through budget management, progressive disclosure (L0-L3 layers), and output constraints. The instructions in SKILL.md and README.md provide structured guidelines for the agent to minimize context usage during search, coding, and debugging tasks without compromising success rates. No malicious behaviors, data exfiltration, or harmful prompt injections were identified; the skill explicitly prioritizes user requests in case of conflict.