Compress subagent prompts into ≤200-word briefs using a five-rule template.
npx skills add https://github.com/LichAmnesia/lich-skills --skill subagent-brief
Please help me install this Agent Skill. Skill: subagent-brief Source: https://github.com/LichAmnesia/lich-skills/tree/main/skills/subagent-brief Command: npx skills add https://github.com/LichAmnesia/lich-skills --skill subagent-brief
Compressing subagent prompts before spawning to minimize token costs and avoid token-bleed in multi-agent workflows.
Spawn a subagent only after generating a ≤200-word brief that follows the five rules and the brief template.
High-intent search queries and answers about installing and using this skill.
To reduce token cost in multi-agent LLM workflows, compress every subagent prompt into a ≤200-word brief using a structured template to strip redundant context and prevent token-bleed.
Subagent prompt compression is a pre-flight discipline that replaces files with paths and context with summaries, generating a concise brief to minimize token waste before launching subagents.
Write a structured subagent brief by applying five rules to strip context, replace files with paths, and compress text into a ≤200-word block containing ROLE, GOAL, INPUTS, CONSTRAINTS, and RETURN sections.
You should avoid using subagents when a fan-out check indicates unnecessary complexity, ensuring you only spawn subagents for orchestration tasks that strictly require distributed processing.
Multi-agent LLM orchestration benefits from strict prompt limits by compressing subagent briefs to ≤200 words, which minimizes token costs and ensures consistent, testable task briefing.
Spawning subagents with full context causes token-bleed because redundant information accumulates across multi-agent calls, making prompt compression necessary to eliminate wasted tokens.
apocohqExtracts structured meeting summaries from VTT transcripts for AI context loading.
vinhnxvCoordinate distributed multi-agent workflows using file-based handoffs and inscription-driven protocols.
distribucionesfayosmoreno-sysOrchestrate distributed system workflows using Temporal saga, entity, and fan-out/fan-in patterns.
baphledStructure prompts with explicit intent and reduced noise to optimize token usage.
